The Effects of Intelligent Semantic Analysis Techniques on Language Acquisition in the Improvement of English Intercultural Communication Skills

With the deepening of globalization and the continuous promotion of the “One Belt, One Road” initiative, China is increasingly in need of international talents with cross-cultural communication skills. It has become an important mission of Chinese education to cultivate a new generation of talents who have a global vision, know international rules and are able to participate in international affairs and international competition [1-4]. At present, the importance of intercultural communicative competence is becoming more and more prominent in university English teaching. Cultivating students’ intercultural communicative competence has become an important task in university English teaching, and intelligent semantic analysis technology plays an important role in intercultural communicative teaching activities [5-8].

Semantic analysis technology refers to the conversion of linguistic information in texts into machine-understandable forms so that computers can understand the actual meaning of sentences and documents [9-10]. Semantic analysis technology is a major branch of the field of artificial intelligence, which utilizes natural language processing (NLP) technology to convert natural language text into meaningful machine-readable information, thus helping computers to understand the real meaning of the text [11-14]. Traditional text processing techniques can only simply process and match the text, and cannot understand the meaning. Semantic analysis technology, on the other hand, can transform text information into a form that can be understood by computers, thus realizing in-depth analysis and understanding of text information [15-18].

The study builds an AMR intelligent semantic analysis system based on stack-LSTM algorithm. After building the system database, the similarity algorithm is used to find semantic differentiation, and then the label conversion algorithm and graph-to-graph linear transformation algorithm are proposed. In order to verify the role of AMR intelligent semantic analysis system in the enhancement of intercultural communicative competence in English language acquisition, an experimental environment was built and realized. Firstly, a comparison with other syntactic analysis tools is made, which is used to compare out the accuracy of the English and Chinese languages, and then the corpus’s is subjected to the experiment of identifying the accuracy of the utterance labeling, and finally the actual effect is analyzed.

2

Intelligent Semantic Analysis in Intercultural Communication Competence

2.1

Semantic Analysis in the Context of English Intercultural Communication

English in cross-cultural context may have different meanings in different occasions, so we need to take into account the cross-cultural communication context when analyzing the semantics of English, and understand the semantics of these words and phrases in the context of different cultures.

2.1.1

Phrases

Many phrases in English often have great differences between their literal meanings and their actual meanings, which is the most typical problem that needs to be noticed in semantic analysis of English in the context of cross-cultural communication. The semantic analysis of English cannot be completely translated directly, but needs to take into account the rules of language use in the cross-cultural context, that is, to consider the special meaning of the phrases in actual use.

2.1.2

Sentences

Sentences consist of various phrases and some auxiliary words. Understanding the semantics of English in the context of cross-cultural communication, especially when facing some English utterances, actually the core lies in grasping the core vocabulary and phrases in the sentences, and understanding the meaning of the whole sentence through the actual meanings of these vocabulary and phrases.

2.1.3

Euphemisms

Euphemisms are also a very common way of expression in English. Although many foreigners prefer to express themselves directly in terms of language communication, they sometimes use euphemisms to express themselves at certain times to avoid embarrassment or to be polite, taking into account some historical and religious factors.

2.2

Cross-language semantic analysis process

Lexical semantic analysis has great advantages, the most prominent of which is that it no longer fixes each word as a set of combinations of meanings, but breaks it down into a number of meanings and puts these meanings into different contexts for the learners to understand and use, which can improve the rote learning habits of students and stimulate the students’ exploration of and love for the language. The structure of the intelligent algorithm for English semantic analysis of utterance components is shown in Figure 1. As can be seen in Figure 1, the structure of the intelligent algorithm based on semantic analysis firstly needs to decompose the English words contained in a sentence into individual words, then retrieve the lexical properties and the corresponding lexical meanings of each word, and finally analyze the meaning of the word in the sentence in context.

2.2.1

System database

The system database is based on a collection of teachers and textbooks, with practitioners compiling the vocabulary that students need to learn based on their different levels of study and syllabus. It mainly contains tables with basic vocabulary information and more specific vocabulary information. Basic Vocabulary Information: According to the syllabus, information such as the lexical nature of the required vocabulary and the pronunciation phonetic symbols in different environments are stored in the database. Special vocabulary information: In English, most nouns have certain formation rules for their plurals, verbs for their past tenses and past participles, and present tenses, adjectives for their comparatives and superlatives, etc. However, there are some vocabulary words for which the above information is inconsistent with the common ones, and they have special characteristics. In order to realize semantic analysis intelligent algorithms more precisely, such words should be stored separately in the database.

2.2.2

Cross-language dependent syntactic analysis

1)

Lexical analysis and lexical analysis

Lexical analysis is the process of taking apart each word in a sentence individually and analyzing its lexical nature, lexical meaning, etc. The lexical analysis of utterance is mainly based on lexical analysis, which first obtains the lexical nature and lexical meaning of each word, and then, based on the obtained data and the lexical nature of the words before and after the word, it determines the lexical nature of the word that the word plays in this utterance.

2)

Representation learning method for language independent features

The key problem that needs to be solved in cross-language dependent syntactic analysis is the variability of the two languages themselves, the source language and the target language, both at the word level and at the sentence structure level. This variability is mainly due to the fact that we use different words to express the same meaning in different languages, and each language has its own grammar to constrain the order of occurrence of words [19]. The representation learning method for language independent features is to formulate the same feature rules among different languages to represent words from multiple languages in the same feature space. Linguistically independent feature representation learning is mainly used to compensate for the differences between languages and establish inter-language connections by inducing the generation of linguistically independent features, and then train the dependent syntactic analyzer through the vector representation of the source language in the feature space, so that the dependent syntactic analyzer can carry out the dependent syntactic analysis of the target language.

Each word pair pair = 〈E_i,F_i〉 in the set of bilingual word pairs V represents the equivalence of English word E_i and Chinese word F_i. After constructing the set of bilingual word pairs V, we will learn to obtain a vector representation for each word pair Pair in the set V by utilizing the deep neural network approach. This vector representation is capable of representing any of the words in the word pair simultaneously. The model used for the learning process is shown in Figure 2.

The input layer of the model shown in Figure 2 is introduced first. From the corpus of the source language, a clause X m is selected as the representation of word w_i centered on word W_i by a specified window length C. A window length of C means to w_i must be centered on W₁W₂......W_c. Then for each word W_i in the words X obtain a representation of the word direction of the word, denoted as R(W_i) . Splice each R(W_i) together m to form a R(X) matrix as in equation (1): (1) $R (X) = [R (W_{1}); R (W_{2}); R (W_{3}) \dots \dots R (W_{c})]$

After obtaining the R(X) matrix, each word in clause X has a corresponding vector representation in the R(X) matrix. The R(X) matrix can then be seen as a representation of clause X. We use R(X) as the input to the deep neural network model. Next, the neural network model will be based on the current input R(X) and the output obtained by nonlinearly transforming the input R(X) using the tanh function of the hidden layer of Fig. 2 is H₁(X), as shown in (2): (2) $H_{1} (X) = \tanh (W_{1} \times R (X) + b_{1})$

where W₁ is the weight of matrix R(X) and b₁ is the bias. H₁(X) is the output vector of the first hidden type. From the model in Fig. 2, it can be seen that the output of each hidden layer will be used as the input of the hidden type behind it, which is then transformed by a nonlinear transformation to get a new output again. The exact number of hidden layers is set dynamically in: the experiment, for example, the output of the i th hidden is H_i(X) and its input is H_i–1(X), as shown in (3): (3) $H_{i} (X) = \tanh (W_{i} \times H_{i - 1} (X) + b_{i})$

Similarly, where w_i is the weight matrix of the current layer, b_i is the bias, and H_i(X) is the output of the i th hidden layer. Once the number of hidden layers t in the model is determined, the output of the last layer is needed to calculate a final score S(X).

(4)

S (X) = θ \times H_{t} (X) + μ

The parameters included in equations from (1) to (4) are the hidden layer parameter ${W_{i}, b_{i}}_{i = 1}^{t}$ and the final output layer parameter (θ,μ). The above is the representation process of word vectors, and the next is the parameter solving process. In the representation learning process we usually utilize the binary classification problem to solve the parameters. First, construct the positive negative class training set $D = {x_{i}, x_{i}^{-}}_{i = 1}^{N}$ , where x_i = W₁W₂......W_c, i.e., the clauses extracted by a fixed window length C, are used as positive samples. And its negative sample $x_{i}^{-}$ is obtained by replacing the center word of x_i with the word paired with x_i in the set of bilingual word pairs V. We input each of the positive and negative samples in D into the model in Fig. 2 and obtained two scores S(x_i) and $S (x_{i}^{-})$ . For solving the parameters we defined the loss function as shown in (5): (5) $J (D) = \frac{1}{N} \sum_{i = 1}^{N} (0, 1 - S (x_{i}) + S (x_{i}^{-}))$

The stochastic gradient descent method is utilized to solve the above equation, and the values of the model parameters are continuously updated over several iterations. By training the above model we obtain the word embedding matrix R for all words in both languages, and each word pair in the set of bilingual word pairs V shares the same language-independent word vector representation R(W_i). In order to make the dependency syntactic analyzer trained on the source language side work on the target language side, we train a non-lexicalized dependency syntactic analyzer by replacing the specific words in the English sentences on the source language side by the international common lexical tags corresponding to each word together with its word vector representations R(W_i) as the input.

2.2.3

Semantic similarity calculation

The vector space modeling criterion is used in measuring utterance similarity. Vector space modeling involves separating the smallest semantic units such as words and phrases in a text and using their calculated similarity as vector elements. Teaching cosine is used in two English sentences to obtain the inter-semantic similarity.

In the vectorized representation of English utterances, two utterances are first represented as equal length vectors, e.g., for utterances T₁ and T₂, all the words of the two utterances are assembled into a joint word set T as in equation (6).

(6)

T = T_{1} \cup T_{2} = {w_{1}, q_{1}, \dots, w_{m}, q_{n}}

Removing the same words in T₁ and T₂ ensures the mutual dissimilarity of the elements in the union set T, where {w₁,w₂,…w_m} is the set of words in statement T₁ and {q₁,q₂,…,q_n} is the set of words in statement T₂, e.g., for English statements:

T₁:{What is your favorite sports?}

T₂:{What kind of sports do you enjoy most?}

Combine the two statements by removing the articles and exclamations from both statements, keeping the real word prototypes, and recording the same words to obtain the combined statement T: {What is your favorite sport kind of do you enjoy most?}. Represent the joint utterance T as a vector S. The length of the words of the joint semantic vector is the same as the number of joint utterances, and at the same time, the utterance T₁ is denoted by asking the joint semantic vector S₁, and T₂ is denoted by asking the joint semantic vector S₂. The words in the vector are in the form of component values, and S_i is taken to be 1 if S_i is included in the semantic vector, and the similarity is computed according to the equation (1) if it is not included in the utterance. Obtain semantic vectors S₁ and S₂ corresponding to utterances T₁ and T₂.

(7)

\begin{array}{l} S_{1} : {1, 1, 1, 1, 1, 0.8, 0.8, 0, 0, 1, 0.8, 0.9} \\ S_{2} : {1, 0, 1, 0.8, 1, 1, 1, 1, 1, 1, 1} \end{array}

where the decimal obtained from the calculation is the value of similarity corresponding to the word. The prepositions and auxiliary verbs such as of, do, etc. in the utterances have no comparative words. After determining the semantic vectors S₁ and S₂ corresponding to statements T₁ and T₂, the similarity of T₁ and T₂ is calculated according to Eq. (7), and the alternative answer with a similarity value greater than the set threshold is taken as the final answer as in Eq. (8).

(8)

s i m (T_{1}, T_{2}) = \frac{S_{1} \cdot S_{2}}{‖ S_{1} ‖ \cdot ‖ S_{2} ‖}

2.2.4

Cross-Language Annotated Transformation Learning Algorithms

1)

Label transformation algorithm

(1) Generate pseudo-labeled data Firstly, the label transformation probability based on arc alignment is counted from parallel labeled data D_s(C^*) and D_t(C^*). Given a certain text X, X ∈ C^* , its annotations under the source and target specifications are D_s(X) and D_t(X), respectively. Suppose (i,j,l) denotes a dependent arc in text X pointing to the i th word from the j th word, and its semantic label on the arc is l. In this chapter, firstly, we count the frequency of occurrence of quaternion (p_d,p_h,l_s,l_t) for all arcs (i.e., (i,j,l_s)∈D_s(X) and (i,j,l_t)∈D_t(X)) appearing in the source and target specifications at the same time. Where p_d, p_h denote the lexical propertie s of the child and parent words in this arc, while l_s, l_t denote the semantic labels of this arc under the source and target specifications, respectively. Thus the final label transformation probability is: (9) $P (l_{t} | p_{d}, p_{h}, l_{s}) = \frac{C o u n t (p_{d}, p_{h}, l_{s}, l_{t})}{\sum_{t'}, ε_{t} C o u n t (p_{d}, p_{h}, l_{s}, l_{t'})}$

(2) Two-step training the training data used for the first training step is the pseudo-labeled data of the target specification obtained through the previous method. In order for the analyzer to be better adapted to the real data distribution, the training data used for the second training step (fine-tuning) is the manually labeled parallel corpus D_t(C^*).

2)

Graph-to-Graph Linear Transformation Algorithm

Unlike the label transformation algorithm, the graph-to-graph linear transformation algorithm directly learns a linear function that transforms the analyzer trained on the source canonical data to the target canonical. Since the Biaffine attention matrix is a core component of the Biaffine parser, it contains important information for making predictions about semantic dependency graphs. Therefore a natural approach is to inherit the parser trained on the source specification, which in turn helps to train the parser on the target specification.

Specifically, assume that U_s, W_s and b_s are the parameters in the source analyzer, respectively. The two linearly transformed functions V_u and V_w act on 6 U_s, W_s, respectively, which in turn give the parameters U_t, W_t of the analyzer under the target specification.

(10)

U_{t} = U_{s} * V_{u}

(11)

W_{t} = W_{s} * V_{w}

The final target analyzer is: (12) $B i a f_{t} (x_{i}, x_{j}) = x_{i}^{T} U_{t} x_{j} + W_{t} (x_{i} \oplus x_{j}) + b_{s}$

2.2.5

Stack-LSTM networks

LSTM is widely used in various tasks in the field of natural language processing, and its structure based on recurrent neural networks allows the model’s output at a certain positional moment to carry information with simultaneous context [20]. The network structure of LSTM is similar to that of RNN, in that it is a repetitive module connected in a recurrent manner, and unlike the RNN, the LSTM has four internal neural network layers that interact in a very specific way, and its internal network structure is shown in Figure 3.

The LSTM has three structures called “gates”, which consist of a sigmoid neural network layer and an element-by-element multiplication operation, which represents a method of selectively letting information pass through, the element-by-element multiplication means multiplying the input by a number in the interval 1, where 0 means it cannot pass through, and (0,1) means it passes through all of them.

The key to the LSTM is the cell state C_t, which in the diagram is the horizontal line running across the top of the diagram, it represents the state of the input sequence up to the current point in time of the network, it goes through two element-by-element operations at each repetition of the module, the first element-by-element multiplication is done to select which positions in the previous state need to be forgotten and which need to be remembered; the second element-by-element addition will determine what new information will be added to the cell state [21]. The formula here is shown in (13): (13) $C_{t} = f_{t} \otimes C_{t - 1} \oplus i_{t} \otimes {\tilde{C}}_{t}$

In the above equation, ${\tilde{C}}_{t}$ represents the new information, i_t represents the filtering of the new information and they are calculated as shown in Equation (14) and Equation (15), x_t is the input at the current point in time, σ is the sigmoid function, which is calculated as shown in (16), and the output is the value in the interval of (0,1), which is used for filtering the information: (14) $i_{t} = σ (W_{i} [h_{t - 1}; x_{t}] + b_{i})$ (15) ${\tilde{C}}_{t} = \tanh (W_{c} [h_{t - 1}; x_{t}] + b_{c})$ (16) $σ (x) = \frac{1}{1 + e^{- x}}$

Finally the network will determine the value of the output based on the cell state, the output of the previous time point and the input of the current time point, calculated as shown in Eqs. (17) and (18), where o_t denotes the result of the output gate, and finally the final hidden output h_t is obtained by doing an element-by-element multiplication with the cell state C_t: (17) $o_{t} = σ (W_{o} [h_{t - 1}; x_{t}] + b_{o})$ (18) $h_{t} = o_{t} \otimes \tanh (C_{t})$

The stack-LSTM adds a stack pointer to the normal LSTM. similar to the traditional LSTM, new inputs are always added to the last position, but when computing the new memory part, the stack pointer determines from which position of the LSTM to provide the c_t–1 and h_t–1 needed for the computation. In addition to adding new elements to the end of the sequence, the stack-LSTM provides the action of bouncing the stack by pointing the stack pointer to the previous element. Thus, this stack-LSTM can be viewed as a stack, where the press-stack action places the new element at the end of the sequence, saves a pointer to the current top of the stack, and then points the top-of-stack pointer to the new element, and the pop-stack action simply updates the top-of-stack pointer without performing additional operations. The output of the layer pointed to by the top-of-stack pointer is a vector representation of all the contents of the stack in continuous space.

2.3

Stack-LSTM-based semantic analysis of AMR

1)

word representation

Bidirectional LSTMs, in contrast to uni-LSTMs, can encode information about the current position and before, in addition to information about the current position and after through LSTMs in the other direction. More formally, we combine two vectors as the input representation of the word, a character-based word representation ${\tilde{w}}_{C}$ and a pre-trained word vector representation ${\tilde{w}}_{L M}$ that does not participate in the training update, and connect the two through a parameter matrix V with a bias term b as the final input representation, computed as shown in Eq. (19): (19) $x_{t} = \max {0, V [{\tilde{w}}_{c}; {\tilde{w}}_{L M}] + b}$

2)

Transfer system state representation

In this paper, three stack-LSTMs are used to learn the representations of the three stacks σ, δ, and β in the above list-based system s =(σ,δ,β,A) respectively, and an additional stack-LSTM is used to learn the representation of the system’s history of transferring action sequences A. At the moment t, the representation of the state e_t of the whole transfer system is shown in Eq. (20): (20) $e_{t} = \max {0, W [s_{t}; d_{t}; b_{t}; a_{t}] + d}$

where W is the parameter matrix, s_t, d_t, b_t, and a_t are the encoding of stack-LSTM at time point t for σ, δ, β, and A, respectively, and d is the bias term, using ReLU as the activation function for the output.

3)

Predicting transfer actions

Our model uses the transfer state e_t at time point t to compute the probability distribution of performing a transfer action in that state, as shown in Equation (21): (21) $p (z_{t} | e_{t}) = \frac{\exp (g_{z_{t}}^{T} e_{t} + q_{z_{t}})}{\sum_{z' \in A} \exp (g_{z_{z}}^{T} e_{t} + q_{z'})}$

In the above equation g_z is a column vector for the representation of transfer action z and q_z is a bias term for transfer action z.

4)

Predicting CONFIRM action generation nodes

Once the model has selected the transfer action CONFIRM, the model needs to determine the nodes in the AMR graph generated from the top word of the buffer, again by categorizing state e_t as shown in Equation (22): (22) $p (c_{t} | e_{t}) = \frac{\exp (g_{c_{t}}^{T} s_{t} + q_{c_{t}})}{\sum_{c' \in N} \exp (g_{c'}^{T} s_{t} + q_{c'})}$

In the above equation N represents all possible candidate concept nodes for the top element of the buffer, g_e is a column vector for the vector representation of concept node c, and q_e is a bias term. For the possible concept nodes generated for each word, we use the training data from AMR to perform the statistics without using additional resources.

The softmax classification in Eq. (21) and Eq. (22) is applied on the state representation of the transfer system e_t represented by the model, and by performing the series of operations described above, we obtain a sequence of transfer actions as the output of the final model, which is then used to generate an AMR graphical representation of the transfer sequence.

3

Intelligent Semantic Analysis Applications in Language Acquisition

3.1

Syntactic standardness analysis of intelligent semantic technologies

3.1.1

Analysis of syntactic labeling accuracy

The most basic guarantee of confidence in language research depends on the accuracy of the annotation. In this study, the accuracy of dependency syntactic structures is evaluated in comparison with Mate Parser, AMR, and Malt Parser intelligent techniques, and the results are shown in Table 1.

Table 1.

The accuracy of the syntax analysis tool is calculated

Analyser	Language	Style	Dependent relation			Whole sentence
Analyser	Language	Style	Mini	Max	Average
AMR	English	Literature	15.88%	100%	87.56%	41%
	English	Inliterature	37.28%	100%	86.91%	34%
	Chinese	Literature	11.4%	100%	81.42%	32%
	Chinese	Inliterature	32.55%	100%	84.51%	23%
Mate Parser	English	Literature	27.12%	100%	86.31%	28%
	English	Inliterature	13.21%	100%	85.33%	21%
	Chinese	Literature	10.2%	100%	77.51%	16%
	Chinese	Inliterature	45%	100%	80.55%	7%
Malt Parser	English	Literature	13.21%	100%	80.78%	28%
	English	Inliterature	40%	100%	81.21%	20%
	Chinese	Literature	11.4%	100%	76.57%	17%
	Chinese	Inliterature	41.51%	100%	78.32%	6%

There is a large gap in the accuracy of AMR semantic tool compared to the lexical assignment tool, both for the annotation of dominant nodes and dependencies, the average accuracy of the syntactic parser ranges from 76% to 86%, while the highest fully correct rate for the whole sentence is only 41%. In addition, the data shows that language and genre also affect the accuracy of syntactic analysis. Among them, the syntactic annotation accuracy of English is significantly higher than that of Chinese, both in terms of local dominant nodes and dependencies and whole sentence accuracy, probably due to the fact that Chinese does not have a rich morphological tagging system, and lexical assignment based on morphological analysis is one of the important reference indexes of the syntactic analyzer.

3.1.2

Error analysis of syntactic labeling

The automatic syntactic analysis errors of English and Chinese were comparatively analyzed using AMR semantic technology, and the results are shown in Table 2 and Table 3, respectively. As can be seen from Tables 2 and 3, the errors of each analyzer show certain commonalities: the analysis of sentence structure has more errors than the analysis of phrase structure, and stack-LSTM has the least number of errors, which is significantly lower than that of Mate Parser and Malt Parser. Compared with English, the number of syntactic analysis errors in Chinese is significantly higher, but the types of errors are somewhat consistent with the results of English analysis.

Table 2.

Error analysis of English sentence collection

Analyser		AMR			Mate Parser			Malt Parser
Syntax type	Grammatical type	Lexical error	Dependent error	Microscope	Grammatical type	Lexical error	Dependent error	Grammatical type	Lexical error	Dependent error
Phrase structure	Modification relation	15	127	182	26	145	218	27	199	284
Phrase structure	Functional relation	5	35	182	9	38	218	2	56	284
Sentence structure	Component relation	8	120	261	19	151	331	18	227	467
Sentence structure	Small sentence relation	9	124	261	18	143	331	13	209	467
Other	Undefined dependencies	1	7	8	1	15	16	1	11	12
Total		38	413	451	73	492	565	61	702	763

Table 3.

Error analysis of Chinese sentence collection

Analyser		AMR			Mate Parser			Malt Parser
Syntax type	Grammatical type	Lexical error	Dependent error	Microscope	Lexical error	Dependent error	Microscope	Lexical error	Dependent error	Microscope
Phrase structure	Modification relation	25	78	162	27	145	214	121	111	311
Phrase structure	Functional relation	11	48	162	1	41	214	29	50	311
Sentence structure	Component relation	42	185	358	69	197	462	72	218	430
Sentence structure	Small sentence relation	46	85	358	41	155	462	45	165	430
Other	Undefined dependencies	4	22	26	12	23	35	14	31	45
Total		128	418	546	150	561	711	281	575	786

3.2

Intelligent Semantic Technology for Recognizing Linguistic Analysis

The experiment uses intelligent semantic techniques to process the collected sentence set. The sentence set has a total of 8000 interrogative sentences. 5367 sentences were randomly selected as the training corpus, and then 1535 interrogative sentences were arbitrarily selected from the remaining sentences as the test corpus.

According to the characteristics of the sentences, many semantic components will not appear or appear with low frequency. Since the feature selection in the experiment is rich and the feature space is huge, pruning before selecting features greatly improves the performance of the system. In this experiment, the features with feature occurrence frequency above times are trained, in which the pruned components are giving things, causing, object, than things, sub things, basis, materials, conditions, indicating judgment components, and other components. The number of features and predictions for training corpus I and the test corpus are shown in Table 4 below.

Table 4.

Quantity statistics

Semantic component	Training language		Test language		Semantic component	Training language		Test language
Semantic component	Characteristic number	Forecast number	Characteristic number	Forecast number	Semantic component	Characteristic number	Forecast number	Characteristic number	Forecast number
Tie	7981	6621	256	242	Degree	567	531	19	15
Consul	26	25	1	1	Range	4632	3765	134	108
Suffer	25	13	1	0	Trend	1999	1891	71	69
Results	821	710	26	23	Cause	30	27	1	1
Guest	5632	4982	155	141	Purpose	26	26	1	1
And things	172	132	5	4	Predicate	7255	7168	256	242
Party	24	11	0	0	Marker	5176	5133	186	185
Tools	246	231	7	6	Unit	156	156	6	6
Mode	322	315	10	9	yw	79	67	2	2
Space	5834	5521	198	182	ots	3978	3786	135	126
Time	166	121	5	4	-	-	-	-	-

The experimental training results of each semantic component labeling evaluation are shown in Fig. 4, Fig. 5 and Fig. 6. As can be seen from the figures, the recognition accuracy of the department, consul, result, guest, tool, way, place, degree, tendency, reason, and purpose is above 92%, with good recognition effect. The recognition accuracy with matter, time, and scope is 82% to 91%, and the recognition effect is average, while the recognition effect with subject matter and parties is poor.

3.3

Analysis of the Impact of Intelligent Semantic Analysis Technology on Language Acquisition

In this section, 100 students from senior and junior levels of an English major class in a university were selected for regression analysis of the influence of using and not using intelligent semantic analysis technology on the students’ language acquisition, and for the accuracy of the results of the study, there were 50 students who used the intelligent semantic technology, and 50 students who did not use it. The results of the analysis are shown in Table 5. From the table, it can be seen that: the effect of whether or not to use intelligent semantic analysis technology in high and low grades on students’ school language use status is shown as follows: whether or not to use intelligent semantic analysis technology in high grades does not have a significant effect on students’ school language use status; whether or not to use intelligent semantic analysis technology in low grades does not have a significant effect on students’ school language use status, either; it is not clear whether or not to use intelligent semantic analysis technology in low grades is not a significant effect on students’ school language use status in comparison with no using intelligent semantic analysis technology compared to exp (1.774) = 5.86, significant at the 0.1 level, which means that after controlling for other factors, students who were not sure whether they would use intelligent semantic analysis technology in the lower grades were 4.7 times more likely to use intelligent semantic analysis technology (compared to not using intelligent semantic analysis). And there is no significant difference in the comparison between using both and not using intelligent semantic analysis technology, exp (.965) = 2.58. The effect of whether or not teachers use intelligent semantic analysis technology in their lessons on students’ language acquisition status is shown in the following way: after controlling for other factors, students whose teachers use intelligent semantic analysis technology in their lessons (compared to those whose teachers do not) are more likely to use it in their lessons (compared to those whose teachers do not). using intelligent semantic analysis technology (compared to not using intelligent semantic analysis technology) were 1.04 times more likely to acquire language, with a significant difference of exp (.707) = 2.01, while the difference between using both modalities (compared to not using intelligent semantic analysis technology) was not significant, with a difference of exp (.225) = 1.23. The effect of gender on students’ language acquisition status was demonstrated by the fact that after controlling for other factors , the odds of using intelligent semantic analysis technology (over not using intelligent semantic analysis technology) between male students and their classmates decreased by 41% compared to female students, exp (-.577) = 0.54, significant at the 0.05 level. After controlling for other factors, the odds of using both technologies between male students and their classmates (over not using intelligent semantic analysis technology) decreased by 29% compared to female students, exp (-.375) = 0.69, significant at the 0.1 level.

Table 5.

The student school language USES regression analysis

	Variable assignment	Regression coefficient	Standard error	Z value	Significance
Use and do not use
Seniors will use	no =0
	yes = 1	.06429	.4183	0.13	0.996
	Inclarity =2	-.9618	1.0717	-0.81	0.289
Lower grades will use	no =0
	yes = 1	-.3649	.3916	0.9	0.371
	Inclarity =2	1.7742	0.9853	1.77	0.002
Teachers don’t use it	yes=0	.7065	.2382	2.93	0.023
Gender	Female=0	-.5794	.2332	2.45	0.033
Age		-.4206	.0892	-4.66	0
Constant term		5.4251	1.255	4.29	0
Both are involved (use and not use) VS Not use
Seniors will use	no=0
	yes=1	-.2769	0.3868	-0.69	0.593
	Inclarity = 2	.0109	.9797	0.02	0.997
Lower grades will use	no=0
	yes=1	.3011	.3652	0.8	0.396
	Inclarity =2	.9651	.9485	1.06	0.314
Teachers don’t use it	yes=0	.2249	.2293	0.95	0.341
Gender	female=0	-.3751	.2248	-1.63	0.095
Age		-.1554	0.0877	-1.74	0.082
Constant term		2.638	1.235	2.11	0.153
N	578
LRchi2	59.58
Prob>chi2	0.0000
Log likeihood	-588.6723
Pseudo R2	0.0402

4

Conclusion

This study proposes a stack-LSTM-based AMR intelligent semantic analysis method to achieve intercultural communication competence in English language acquisition. The intelligent semantic system is capable of analyzing syntax with good accuracy and a low error rate. Through the semantic component annotation recognition experiment, the AMR intelligent semantic system is used for recognition, and the experimental results show that the intelligent semantic system proposed in this paper achieves an accuracy of 92% for semantic component annotation recognition, and the annotation recognition effect is good. The intelligent semantic analysis system is used in practice to analyze the influence of language acquisition brought by the system through the use of different types of people, and the results show that the difference in the enhancement of their own language communication ability is significant between the teachers’ population who use the intelligent semantic system for teaching and the students who do not use the intelligent semantic system for teaching.

Idioma:: Inglés

Calendario de la edición:: 1 veces al año
Temas de la revista:: Ciencias de la vida, Ciencias de la vida, otros, Matemáticas, Matemáticas aplicadas, Matemáticas generales, Física, Física, otros

RSS Feed de revista

The Effects of Intelligent Semantic Analysis Techniques on Language Acquisition in the Improvement of English Intercultural Communication Skills

Xiangming Huang

Publicado en línea: 17 mar 2025

Recibido: 17 nov 2024

Aceptado: 18 feb 2025

DOI: https://doi.org/10.2478/amns-2025-0264

Palabras claveIntercultural communication, Syntactic analysis, Semantic technology, Stack-LSTM algorithm

© 2025 Xiangming Huang, published by Sciendo

This work is licensed under the Creative Commons Attribution 4.0 International License.

Palabras clave
Intercultural communication, Syntactic analysis, Semantic technology, Stack-LSTM algorithm