Evaluating the random representation assumption of lexical semantics in cognitive models

A common assumption implicit in cognitive models is that lexical semantics can be approximated by using randomly generated representations to stand in for word meaning. However, the use of random representations contains the hidden assumption that semantic similarity is symmetrically distributed across randomly selected words or between instances within a semantic category. We evaluated this assumption by computing similarity distributions for randomly selected words from a number of well-known semantic measures and comparing them with the distributions from random representations commonly used in cognitive models. The similarity distributions from all semantic measures were positively skewed compared with the symmetric normal distributions assumed by random representations. We discuss potential consequences that this false assumption may have for conclusions drawn from process models that use random representations.     

Accessing short-term memory with semantic and phonological information: A time-course analysis

The time course for accessing short-term memory representations on the basis of item, phonological, and semantic information was measured with a speed-accuracy tradeoff (SAT) variant of a recognition probe task. Subjects studied 5-word lists and then judged whether a test word (1) was in the study list (item judgment); (2) rhymed with a word in the study list (rhyme judgment); or (3) was a synonym of a word in the study list (synonym judgment). Analysis of the SAT retrieval functions demonstrated that judgments based on phonological and semantic information were slower and less accurate than judgments based on composite item representations. The slower retrieval dynamics for rhyme and synonym judgments suggest that judgments based on component properties of the memory trace are mediated by either a generate or recall process distinct from the parallel or direct-access mechanism used in item judgments.     

Examining the role of word identification in synonymy judgment

This paper examines the role of word identification in semantic judgment tasks. We argue that word-identification processes can substantially alter the functional properties of semantic judgment tasks. We investigate this empirically by examining the effects of semantic relatedness and relatedness proportion on synonymy judgments of non-synonym pairs. The results demonstrate that semantic relatedness slows synonymy judgments for non-synonym pairs when the relatedness proportion is low, but speeds them when the relatedness proportion is high. The former effect is a replication of traditional findings, while the latter effect reverses these findings. We explain these results by hypothesizing that increasing the relatedness proportion increases the beneficial effects of semantic relatedness in word identification, and this produces the reversal of traditional findings. We close by emphasizing that models of semantic judgment tasks can become more comprehensive by focusing on word identification.     

The roles of semantic similarity and proactive interference in the word length effect

We examined the influence of semantic similarity and proactive interference (PI) on the word length effect (WLE) in immediate serial recall. Word length was manipulated by comparing memory for monosyllabic versus multisyllabic words. PI effects were evaluated by manipulating semantic similarity in the to-be-remembered lists and examining its impact on the WLE’s magnitude across eight-trial blocks. Words were sampled from a single semantic category across the entire block, from a single category within the list, or from different categories. Robust WLEs were observed in single-category blocks and when words were from different categories. However, when all the within-list words were from the same semantic category, the WLE was sharply attenuated. Except for the within-list semantic similarity condition, there was a buildup in PI levels in the form of protrusion errors across trials. However, the magnitude of the WLE did not increase with the PI buildup, suggesting that it was not affected by PI across trials.     

WINDSOR: Windsor improved norms of distance and similarity of representations of semantics

Lexical co-occurrence models of semantic memory form representations of the meaning of a word on the basis of the number of times that pairs of words occur near one another in a large body of text. These models offer a distinct advantage over models that require the collection of a large number of judgments from human subjects, since the construction of the representations can be completely automated. Unfortunately, word frequency, a well-known predictor of reaction time in several cognitive tasks, has a strong effect on the co-occurrence counts in a corpus. Two words with high frequency are more likely to occur together purely by chance than are two words that occur very infrequently. In this article, we examine a modification of a successful method for constructing semantic representations from lexical co-occurrence. We show that our new method eliminates the influence of frequency, while still capturing the semantic characteristics of words.     

Semantic Memory Search and Retrieval in a Novel Cooperative Word Game: A Comparison of Associative and Distributional Semantic Models

Considerable work during the past two decades has focused on modeling the structure of semantic memory, although the performance of these models in complex and unconstrained semantic tasks remains relatively understudied. We introduce a two-player cooperative word game, Connector (based on the boardgame Codenames), and investigate whether similarity metrics derived from two large databases of human free association norms, the University of South Florida norms and the Small World of Words norms, and two distributional semantic models based on large language corpora (word2vec and GloVe) predict performance in this game. Participant dyads were presented with 20-item word boards with word pairs of varying relatedness. The speaker received a word pair from the board (e.g., exam-algebra) and generated a one-word semantic clue (e.g., math), which was used by the guesser to identify the word pair on the board across three attempts. Response times to generate the clue, as well as accuracy and latencies for the guessed word pair, were strongly predicted by the cosine similarity between word pairs and clues in random walk-based associative models, and to a lesser degree by the distributional models, suggesting that conceptual representations activated during free association were better able to capture search and retrieval processes in the game. Further, the speaker adjusted subsequent clues based on the first attempt by the guesser, who in turn benefited from the adjustment in clues, suggesting a cooperative influence in the game that was effectively captured by both associative and distributional models. These results indicate that both associative and distributional models can capture relatively unconstrained search processes in a cooperative game setting, and Connector is particularly suited to examine communication and semantic search processes.     

On Judgments of Approximately Equal

Although precision is often important in quantitative judgment, sometimes, it is valuable to recognize that two quantities are roughly the same. Fuzzy‐trace theory suggests that approximately equal judgments rely on gist representations (i.e., meaningful fuzzy categories of quantity). We conducted three experiments to investigate approximately equal judgments with number pairs presented in different formats, both with and without semantic content (breast cancer statistics). In each study, the ratio of the smaller divided by larger number predicted approximately equal judgments. Experiment 1 also examined how knowledge of breast cancer, presentation format (frequencies vs. percentages), and differences in gist comprehension of breast cancer information influence fuzzy equality judgments. As predicted by the fuzzy‐trace theory concept of denominator neglect, approximately equal judgments were more sensitive, as measured by signal detection theory (SDT) analyses, when presented as percentages. In both experiments with breast cancer statistics, people were more likely to judge number pairs approximately equal when they were embedded in sentences about breast cancer, and breast cancer knowledge predicted increased perception of equality, when appropriately consistent with reliable sources, and increased judgment sensitivity. In Experiment 2, a simple intervention focusing on gist meaning increased source‐consistent approximately equal judgments, increased SDT judgment sensitivity, and decreased SDT response bias. In Experiment 3, using number pairs devoid of semantic context spanning four orders of magnitude, we further examined ratio similarity. Overall, more knowledgeable judges and those who better understood the gist of meaningful numbers were more likely to judge literally different numbers as “approximately equal” rather than make precise discriminations that were meaningless. Copyright © 2017 John Wiley & Sons, Ltd.     

The Micro-Category account of analogy

Here, we investigate how activation of mental representations of categories during analogical reasoning influences subsequent cognitive processing. Specifically, we present and test the central predictions of the "Micro-Category" account of analogy. This account emphasizes the role of categories in aligning terms for analogical mapping. In a semantic priming paradigm, a four-word analogy task was compared to two other four-word tasks. Stimuli were identical in all tasks; only the instructions given to participants differed. Participants were instructed to identify analogy relations, category relations, or conventionalized semantic relations in the four-word sets. After each four-word set, a single target word appeared and participants named this word aloud. Target words that referred to category relations in the preceding four-word sets were primed as strongly when participants identified analogies as when participants identified categories, suggesting that activation of category concepts plays an important role in analogical thinking. In addition, priming of category-referent words in the analogy and category tasks was significantly greater than priming of these words when participants identified conventionalized semantic relations. Since identical stimuli were used in all conditions, this finding indicates that it is the activation of category relations, distinct from any effect of basic semantic association, that causes analogical reasoning to prime category-referent words. We delineate how the "Micro-Category" account of analogy predicts these phenomena and unifies findings from diverse areas of research concerning analogical reasoning.     

Distributional Models of Category Concepts Based on Names of Category Members

Cognitive scientists have long used distributional semantic representations of categories. The predominant approach uses distributional representations of category-denoting nouns, such as “city” for the category city. We propose a novel scheme that represents categories as prototypes over representations of names of its members, such as “Barcelona,” “Mumbai,” and “Wuhan” for the category city. This name-based representation empirically outperforms the noun-based representation on two experiments (modeling human judgments of category relatedness and predicting category membership) with particular improvements for ambiguous nouns. We discuss the model complexity of both classes of models and argue that the name-based model has superior explanatory potential with regard to concept acquisition.     

A study of metrics of distance and correlation between ranked lists for compositionality detection

Compositionality in language refers to how much the meaning of some phrase can be decomposed into the meaning of its constituents and the way these constituents are combined. Based on the premise that substitution by synonyms is meaning-preserving, compositionality can be approximated as the semantic similarity between a phrase and a version of that phrase where words have been replaced by their synonyms. Different ways of representing such phrases exist (e.g., vectors (Kiela and Clark, 2013) or language models (Lioma, Simonsen, Larsen, and Hansen, 2015)), and the choice of representation affects the measurement of semantic similarity.We propose a new compositionality detection method that represents phrases as ranked lists of term weights. Our method approximates the semantic similarity between two ranked list representations using a range of well-known distance and correlation metrics. In contrast to most state-of-the-art approaches in compositionality detection, our method is completely unsupervised. Experiments with a publicly available dataset of 1048 human-annotated phrases shows that, compared to strong supervised baselines, our approach provides superior measurement of compositionality using any of the distance and correlation metrics considered.     

Examining lateralized semantic access using pictures

A divided visual field (DVF) experiment examined the semantic processing strategies employed by the cerebral hemispheres to determine if strategies observed with written word stimuli generalize to other media for communicating semantic information. We employed picture stimuli and vary the degree of semantic relatedness between the picture pairs. Participants made an on-line semantic relatedness judgment in response to sequentially presented pictures. We found that when pictures are presented to the right hemisphere responses are generally more accurate than the left hemisphere for semantic relatedness judgments for picture pairs. Furthermore, consistent with earlier DVF studies employing words, we conclude that the RH is better at accessing or maintaining access to information that has a weak or more remote semantic relationship. We also found evidence of faster access for pictures presented to the LH in the strongly-related condition. Overall, these results are consistent with earlier DVF word studies that argue that the cerebral hemispheres each play an important and separable role during semantic retrieval.     

Semantic priming effects on picture and word processing

The effects of semantic priming on picture and word processing were assessed under conditions in which subjects were required simply to identify stimuli (label pictures or read words) as rapidly as possible. Stimuli were presented in pairs (a prime followed by a target), with half of the pairs containing members of the same semantic category and half containing unrelated concepts. Semantic relatedness was found to facilitate the identification of both pictures (Experiment 1) and words (Experiment 2), and obtained interactions of semantic relatedness and stimulus quality in both experiments suggested that semantic priming affects the initial encoding of both types of stimuli. In Experiment 3, subjects received pairs of pictures, pairs of words, and mixed pairs composed of a picture and a word or of a word and a picture. Significant priming effects were obtained on mixed as well as unmixed pairs, supporting the assumption that pictures and words access semantic information from a common semantic store. Of primary interest was the significantly greater priming obtained in picture-picture pairs than in word-word or mixed pairs. This suggests that, in addition to priming that is mediated by the semantic system, priming may occur in picture-picture pairs that results from the overlap in visual features common to the pictorial representations of objects from the same semantic category.     

Exploring What Is Encoded in Distributional Word Vectors: A Neurobiologically Motivated Analysis

The pervasive use of distributional semantic models or word embeddings for both cognitive modeling and practical application is because of their remarkable ability to represent the meanings of words. However, relatively little effort has been made to explore what types of information are encoded in distributional word vectors. Knowing the internal knowledge embedded in word vectors is important for cognitive modeling using distributional semantic models. Therefore, in this paper, we attempt to identify the knowledge encoded in word vectors by conducting a computational experiment using Binder et al.'s (2016) featural conceptual representations based on neurobiologically motivated attributes. In an experiment, these conceptual vectors are predicted from text-based word vectors using a neural network and linear transformation, and prediction performance is compared among various types of information. The analysis demonstrates that abstract information is generally predicted more accurately by word vectors than perceptual and spatiotemporal information, and specifically, the prediction accuracy of cognitive and social information is higher. Emotional information is also found to be successfully predicted for abstract words. These results indicate that language can be a major source of knowledge about abstract attributes, and they support the recent view that emphasizes the importance of language for abstract concepts. Furthermore, we show that word vectors can capture some types of perceptual and spatiotemporal information about concrete concepts and some relevant word categories. This suggests that language statistics can encode more perceptual knowledge than often expected.     

Goal‐Derived Categories: The Role of Personal and Situational Goals in Category Representations

Prior research often emphasized a stimulus‐based or bottom‐up view of product category representations. In contrast, we emphasize a more purposeful, top‐down perspective and examine categories that consumers might construct in the service of salient (i.e., highly accessible) goals. Specifically, we investigate how the point of view imposed by salient consumer goals might affect category representations assessed by participants’ similarity judgments of food products. A key factor in our study is that we examine both individual and situational sources of variability in goal salience. In addition, we also vary the surface‐level, visual resemblance of the stimulus pairs of foods used in the study. The results suggest that personal goals (e.g., health) and situational goals (e.g., convenience) act in conjunction and exert a systematic impact on category representations. Both types of goals, when salient, enhanced the perceived similarity of goal‐appropriate products and reduced the similarity of product pairs when only one product was ideal for the particular goal. The similarity‐enhancing effect was most pronounced when the surface resemblance between the products was low, and the similarity‐diminishing effect was more apparent when surface resemblance was high. Implications are discussed for current theoretical assumptions regarding categorization in consumer research.     

What makes words sound similar?

Although similarity plays an important role in accounts of language processing, there are surprisingly few direct empirical studies of the phonological similarity between words, and it is therefore not clear whether similarity comparisons between words involve processes similar to those involved in other cognitive domains. In five experiments, participants chose which of two monosyllabic pseudo-words sounded more similar to a target pseudo-word. Our results are generally consistent with the structural alignment theory of comparisons between complex mental representations, suggesting that phonological word similarity parallels similarity involving other kinds of information including visual objects and scenes, events, and word meanings. We use our results to test new metrics of word similarity, and identify predictions for future similarity research both in the domain of word sounds and in other cognitive domains.     

Process demands of rejection mechanisms of recognition memory

A surge of research has been conducted to examine memory editing mechanisms that help distinguish accurate from inaccurate memories. In the present experiment, the authors examined the ability of participants to use novelty detection, recollection rejection, and plausibility judgments to reject lures presented on a recognition memory test. Participants studied a list of word pairs that were arranged in a category relationship (both words from the same category) or an unrelated relationship (both words from different categories) under full or divided attention. At test, participants were given a yes/no recognition test in which they were to respond after seeing the test items for 400 ms or 2,800 ms. Some of the test items were rearranged word pairs that were consistent with the study relationship, whereas others were inconsistent with the study relationship. The results demonstrate that the participants required full attention at study to use novelty detection, recollection rejection, and plausibility judgments to reject lures. Moreover, the results indicate that a long response deadline at test was needed for participants to use both recollection rejection and plausibility judgments to reject lures.     

Exemplar by feature applicability matrices and other Dutch normative data for semantic concepts

Features are at the core of many empirical and modeling endeavors in the study of semantic concepts. This article is concerned with the delineation of features that are important in natural language concepts and the use of these features in the study of semantic concept representation. The results of a feature generation task in which the exemplars and labels of 15 semantic categories served as cues are described. The importance of the generated features was assessed by tallying the frequency with which they were generated and by obtaining judgments of their relevance. The generated attributes also featured in extensive exemplar by feature applicability matrices covering the 15 different categories, as well as two large semantic domains (that of animals and artifacts). For all exemplars of the 15 semantic categories, typicality ratings, goodness ratings, goodness rank order, generation frequency, exemplar associative strength, category associative strength, estimated age of acquisition, word frequency, familiarity ratings, imageability ratings, and pairwise similarity ratings are described as well. By making these data easily available to other researchers in the field, we hope to provide ample opportunities for continued investigations into the nature of semantic concept representation. These data may be downloaded from the Psychonomic Society’s Archive of Norms, Stimuli, and Data, www.psychonomic.org/archive.     

The relation between body semantics and spatial body representations

The present study addressed the relation between body semantics (i.e. semantic knowledge about the human body) and spatial body representations, by presenting participants with word pairs, one below the other, referring to body parts. The spatial position of the word pairs could be congruent (e.g. EYE / MOUTH) or incongruent (MOUTH / EYE) with respect to the spatial position of the words’ referents. In addition, the spatial distance between the words’ referents was varied, resulting in word pairs referring to body parts that are close (e.g. EYE / MOUTH) or far in space (e.g. EYE / FOOT). A spatial congruency effect was observed when subjects made an iconicity judgment (Experiments 2 and 3) but not when making a semantic relatedness judgment (Experiment 1). In addition, when making a semantic relatedness judgment (Experiment 1) reaction times increased with increased distance between the body parts but when making an iconicity judgment (Experiments 2 and 3) reaction times decreased with increased distance. These findings suggest that the processing of body-semantics results in the activation of a detailed visuo-spatial body representation that is modulated by the specific task requirements. We discuss these new data with respect to theories of embodied cognition and body semantics.     

Nonmetric scaling of line lengths using latencies,similarity, and same-different judgments

An experiment investigating the relation between various ways of deriving psychophysical functions for line length is reported. A two-phase computer-controlled experiment in which the stimuli were pairs of lines was performed. In the first phase, the pair was presented for several seconds and the S was required to make a category judgment of the similarity of the lines. In the second phase, the pair was presented for a very brief period of time and the S decided whether the lines were the same or of different lengths. Three sets of data were analyzed: (1) the similarity judgments, (2) the proportion of correct different judgments, and (3) the latency of the correct different judgments. An individual differences analysis indicated that there were individual differences in the similarity and latency data. This analysis was not applicable to the proportion data. Nonmetric scaling revealed that the psychophysical functions based on all three sets of data were essentially identical     

语义相似性与类别大小对图-词干扰范式下语义效应的影响

采用基本水平命名和类别水平命名任务, 通过2个实验, 考察了图-词之间的语义相似性与类别大小对图-词干扰范式下语义效应的影响。实验1发现, 图片命名时间在语义相似性高的干扰词条件下显著快于在语义相似性低的干扰词条件下, 但这种效应只出现在基本水平命名任务中, 未出现在类别水平命名任务中。图-词之间的语义相似性高低对低熟悉图片命名影响大, 对高熟悉图片命名影响小。实验2显示, 语义关联的干扰词对图片的基本水平命名产生了语义干扰效应, 对图片的类别水平命名产生了语义促进效应。类别大小强化了语义效应：类别大, 图片的基本水平命名的语义干扰效应更大, 图片的类别水平命名的语义促进效应更强。整个研究表明, 图-词之间的语义关联的强度影响语义效应的方向, 图-词之间的语义关联的广度影响语义效应的强弱。需要结合多种理论来解释图-词之间的语义相似性和类别大小对图-词干扰范式下语义效应的影响。