Learning about the Meanings of Ambiguous Words: Evidence from a Word-Meaning Priming Paradigm with Short Narratives

Fluent language comprehension requires readers and listeners to rapidly select an appropriate meaning for each word that they encounter. This meaning selection process is particularly challenging when low-frequency (subordinate) word meanings are used (e.g. the “river bank” meaning of “bank”). Recent word-meaning priming experiments show that recent experience can help to make subordinate word meanings more readily available, and thereby reduce the difficulty in accessing these meanings. One limitation of previous word-meaning priming experiments is that participants encounter the ambiguous words within a list of unconnected single sentences in which each ambiguous word is strongly disambiguated by words within the prime sentence. The current web-based study (N=51) extends this work to replicate word-meaning priming using short 3-sentence narratives as primes in which relatively weak contextual cues in sentence 1 serve to disambiguate a target ambiguous word that occurs in sentence 3. The results from the subsequent word-association test task confirmed that following a short delay (digit span) task the primed (subordinate) meanings were more readily available compared with an unprimed control. This work represents an important first step in moving the word-meaning priming paradigm towards materials that more reflect the varied ways in which ambiguous words are used within natural language.

Word Sense Disambiguation Using Prior Probability Estimation Based on the Korean WordNet

Supervised disambiguation using a large amount of corpus data delivers better performance than other word sense disambiguation methods. However, it is not easy to construct large-scale, sense-tagged corpora since this requires high cost and time. On the other hand, implementing unsupervised disambiguation is relatively easy, although most of the efforts have not been satisfactory. A primary reason for the performance degradation of unsupervised disambiguation is that the semantic occurrence probability of ambiguous words is not available. Hence, a data deficiency problem occurs while determining the dependency between words. This paper proposes an unsupervised disambiguation method using a prior probability estimation based on the Korean WordNet. This performs better than supervised disambiguation. In the Korean WordNet, all the words have similar semantic characteristics to their related words. Thus, it is assumed that the dependency between words is the same as the dependency between their related words. This resolves the data deficiency problem by determining the dependency between words by calculating the χ2 statistic between related words. Moreover, in order to have the same effect as using the semantic occurrence probability as prior probability, which is used in supervised disambiguation, semantically related words of ambiguous vocabulary are obtained and utilized as prior probability data. An experiment was conducted with Korean, English, and Chinese to evaluate the performance of our proposed lexical disambiguation method. We found that our proposed method had better performance than supervised disambiguation methods even though our method is based on unsupervised disambiguation (using a knowledge-based approach).

Research on Semantic Similarity of Short Text Based on Bert and Time Warping Distance

The research on semantic similarity of short text plays an important role in machine translation, emotion analysis, information retrieval and other AI business applications. However, according to existing short text similarity research, the characteristics of ambiguous vocabularies are difficult to be effectively analyzed, the solution of the problem caused by words order needs to be further optimized as well. This paper proposes a short text semantic similarity calculation method that combines BERT and time warping distance algorithm, in order to solve the problem of vocabulary ambiguity. The model first uses the pre trained Bert model to extract the semantic features of the short text from the whole level, and obtains a 768 dimensional short text feature vector. Then, it transforms the extracted feature vector into a point sequence in space, uses the CTW algorithm to calculate the time warping distance between the curves connected by the point sequence, and finally uses the weight function designed by the analysis, according to the smaller the time warpage distance is, the higher the degree of small similarity is, to calculate the similarity between short texts. The experimental results show that this model can mine the feature information of ambiguous words, and calculate the similarity of short texts with lexical ambiguity effectively. Compared with other models, it can distinguish the semantic features of ambiguous words more accurately.

Training the Interpretation of Ambiguity

<p>The interpretation of emotionally ambiguous words, sentences, or scenarios can be biased through training procedures that are collectively called Cognitive Bias Modification for Interpretation (CBM-I). Little CBM-I research has examined the underlying mechanisms responsible for the induction of emotional interpretive biases, or the potential for this line of enquiry to inform psycholinguistic models of meaning activation and selection. In this thesis a novel CBM-I paradigm was developed and then systematically manipulated in order to discriminate between two mechanistic accounts of changes in interpretive bias – the Emotional Priming Account and the Ambiguity Resolution Account. In Experiment 1 participants completed word fragments that were consistently related to either a negative or benign interpretation of an ambiguous sentence. In a subsequent semantic priming task they demonstrated an interpretation bias, in that they were faster to judge the relatedness of targets that were associated with the training-congruent than the training-incongruent meaning of an emotionally ambiguous homograph. In Experiment 2 the time between the presentation of the prime and the target (Stimulus Onset Asynchrony; SOA) at test was shortened. Interpretive biases were not observed at a short SOA suggesting that training did not induce biases at an early lexical activation stage. Interpretive bias was then eliminated when participants simply completed valenced word fragments (Experiment 3) or completed fragments related to emotional but unambiguous sentences (Experiment 4) during training. Only when participants were required to actively resolve emotionally ambiguous sentences during training did changes in interpretation emerge at test. These findings suggest that CBM-I achieves its effects by altering a production rule that aids the selection of a single meaning from alternatives, in line with an Ambiguity Resolution account. Finally, no interpretive biases were observed when the task in the test phase was substituted with a lexical decision task (Experiment 5). Participants only showed biases in the selection of meanings when the test phase encouraged them to interpret the prime. This pattern of results suggests that the alteration of selection patterns in word recognition depend on the strategies employed by participants in the test phase. Overall the findings are discussed with regards to the Ambiguity Resolution and Emotional Priming accounts of modified interpretive biases. Implications for current psycholinguistic models of meaning activation and selection are considered.</p>

Training the Interpretation of Ambiguity

<p>The interpretation of emotionally ambiguous words, sentences, or scenarios can be biased through training procedures that are collectively called Cognitive Bias Modification for Interpretation (CBM-I). Little CBM-I research has examined the underlying mechanisms responsible for the induction of emotional interpretive biases, or the potential for this line of enquiry to inform psycholinguistic models of meaning activation and selection. In this thesis a novel CBM-I paradigm was developed and then systematically manipulated in order to discriminate between two mechanistic accounts of changes in interpretive bias – the Emotional Priming Account and the Ambiguity Resolution Account. In Experiment 1 participants completed word fragments that were consistently related to either a negative or benign interpretation of an ambiguous sentence. In a subsequent semantic priming task they demonstrated an interpretation bias, in that they were faster to judge the relatedness of targets that were associated with the training-congruent than the training-incongruent meaning of an emotionally ambiguous homograph. In Experiment 2 the time between the presentation of the prime and the target (Stimulus Onset Asynchrony; SOA) at test was shortened. Interpretive biases were not observed at a short SOA suggesting that training did not induce biases at an early lexical activation stage. Interpretive bias was then eliminated when participants simply completed valenced word fragments (Experiment 3) or completed fragments related to emotional but unambiguous sentences (Experiment 4) during training. Only when participants were required to actively resolve emotionally ambiguous sentences during training did changes in interpretation emerge at test. These findings suggest that CBM-I achieves its effects by altering a production rule that aids the selection of a single meaning from alternatives, in line with an Ambiguity Resolution account. Finally, no interpretive biases were observed when the task in the test phase was substituted with a lexical decision task (Experiment 5). Participants only showed biases in the selection of meanings when the test phase encouraged them to interpret the prime. This pattern of results suggests that the alteration of selection patterns in word recognition depend on the strategies employed by participants in the test phase. Overall the findings are discussed with regards to the Ambiguity Resolution and Emotional Priming accounts of modified interpretive biases. Implications for current psycholinguistic models of meaning activation and selection are considered.</p>

Ambiguous Word Processing among Second Language Learners

This study investigated second language learners’ processing of ambiguous words (e.g., bank: [1] a financial institution, [2] an edge of a river/lake) and whether these learners are able to activate the secondary meaning as quickly as they do with the dominant meaning. English L2 and L1 participants used a window paradigm to perform a self-paced reading task, in which all sentences were biased for the secondary meaning (i.e., bank as an edge of a river/lake). The results showed that L1 participants activated both the dominant and the secondary meanings of an ambiguous word even when it is embedded within a secondary-biasing context. However, L2 participants had some difficulty in activating the L2 secondary meaning even when the preceding context was biased for it. The results of the L1 participants were compatible with the autonomous access model in that all meanings of an ambiguous [L1] word are accessed even when the context is biasing for a specific meaning. However, the results of the L2 participants, although they knew both meanings of each ambiguous word in this study based on their post-experiment questionnaire, showed that the more dominant meaning of an L2 ambiguous word is activated first, even when the context is biasing for a secondary meaning.

Language communication in a pragmatic perspective: Flouting the cooperative principle

The aim of the article is a pragmatic analysis of various linguistic communication situations in the light of Grice’s principle of cooperation (1975). The analysis shows that language strategies involve a deliberate flouting of the cooperative principle using various pragmatic functions. The presented communication strategies in English, German, Polish and Russian show similarities in their occurrence. The sender may convey intentions not directly, but by hidden means of expression which often become an exponent of an apparent question, a change in the argumentative direction, the use of ambiguous words, irony or even silence. Hence, we can talk about the implementation of the pragmatic functions of “language avoidance”, “counter-argumentation”, “counter-proposal”, “irony” etc.

Interlocutor modelling in comprehending speech from interleaved interlocutors of different dialectic backgrounds

It has been shown that, in language comprehension, listeners model certain attributes of their interlocutor (e.g., dialectic background, age, gender) and interpret speech against the model; e.g., they understand cross-dialectally ambiguous words such as flat and gas for their American English (AE) meanings more often when listening to an AE interlocutor than a British English (BE) interlocutor. This study reported two experiments to further investigate whether listeners can construct concurrent interlocutor models when communicating with two interlocutors of different dialectic backgrounds and, if they do, how effectively and upon what cues they choose between concurrent models for word interpretation. In Experiment 1, we observed that listeners accessed more AE meanings when listening to an AE than BE interlocutor and such an accent effect that was comparable between listening to blocked interlocutors and listening to interleaved interlocutors. This finding suggests that listeners can construct concurrent interlocutor models and appropriately apply them to constrain word meaning interpretation. Experiment 2 (pre-registered) replicated the finding of concurrent interlocutor models and further showed that listeners chose between concurrent interlocutor models using accent details in a word (such that words pronounced more differently between BE and AE showed a larger accent effect) but not using voice details (such that the accent effect was comparable between listening to interlocutors of the same gender and listening to interlocutors of different genders). In all, our results show that listeners can construct concurrent interlocutor models of dialectic backgrounds and use accent details in a word to determine which model the word should be interpreted against.