Whenever we speak, read or write we always use words which we had to acquire and organize to prevent their forgetting. Yet, none of this guarantees word finding, which is why lexicographers have built dictionaries. Alas, most of them list words in alphabetical order, which generally suites the reader, but hardly ever the writer. Indeed, how can they find in such a resource the form ‘panda’ while their mind sees only a ‘fluffy white bear with black patches around his eyes’?
To address this kind of problem it would be nice to have a lexicon built with the speaker/writer in mind. To achieve this goal, we need to integrate the author into the development cycle right from the start, and to account for diverse research findings regarding:
•
the human brain (modularity, parallelism, dynamism, high connectivity).
•
the mental lexicon: While meanings and forms are presented next to each other in symbolic form in paper dictionaries, they are decomposed and represented subsymbolically in our brain. The word’s critical parts (meaning, form, sound) are distributed across various layers, and their organization at the various levels is by affinity rather than by alphabetical order. These conclusions are clearly supported by empirical data (speech errors, brain scans, etc.). Given these different views concerning words and their organization it is not surprising to see that in one case (traditional dictionaries) words are retrieved as holistic entities, while in the other (brain) they are gradually activated.
•
performance factors affecting
online and
off-line processing, that is, word access in real time (incremental synthesis of word form), and deliberate search in a lexical resource (navigation, retrieval).
•
meta-knowledge: knowledge concerning the organization of the mental lexicon: proximity of source and target word; type of relation between the two, etc.)— and
•
knowledge states, i.e., the knowledge a user has at the onset of the search. Studies concerning the Tip-of-the-Tongue problem have shown that a user being in this state has only access to fragments of the final form, and that these fragments generally pertain to different levels (conceptual, syntactic, phonological).
To be more concrete, we ask how a deeper and broader understanding of the human mind and its search strategies might enable us to develop resources that allow authors to access a desired lexical form via fragmentary input, available at the moment of consultation, though probably activated at different points in time, hence pertaining to different levels (meaning, sound, associated words).
Indeed, lexicographers have already built resources corresponding to the various stages or layers described by psychologists. Yet these resources are not only incomplete, they are also static and not connected to each other. For example, it is extremely rare to see a resource allowing access via meaning ‘and’ sound. Yet, integrating into a single resource information coming from different levels (meaning, form, sound) would allow not only reduce the search space, but also to emulate parallelism. Knowledge of data of one level (say, meaning) may allow us to complete the missing parts of another (sound), hence contribute to the completion of the entire puzzle/picture.
This goal is ambitious, as it poses many problems not addressed in the past. For example, how to reconcile the fact that words are represented holistically in books and computers (dictionaries, electronic resources), while they are decomposed and subsymbolic in the human mind? No doubt, people need symbols at the interface level (input/output of dictionary consultation), and, they are often able to provide only incomplete data (fragments, and data lacking precision), while the machine (or the dictionary) generally expects entire words and precise information.
The scope of this Research Topic is thus two-fold:
1. Consider the lexicon from a cognitive point of view by and large. This is a broad view akin to
CogALex (
https://sites.google.com/view/cogalex-2020/home) standing for
Cognitive Aspects of the Lexicon;
2. Consider the development of an ecosystem to support word finding (word access at the moment of speaking or writing)
This being so, we welcome articles addressing the problems raised in the CogALex workshop series as well as the following points:
(a) Given our goal, and to be fruitful for our discussion, what shall we put behind the term ‘mental lexicon’?
(b) How can we reconcile the fact that words are decomposed, hence subsymbolic, in the human brain, while humans can only provide and interpret symbolic information, which is why we have words in the lexicon?
(c) How to build a representative map of the mental lexicon for a given user group?
(d) Which corpora (knowledge graphs, BabelNet, ConceptNet) and which combinations have the best potential to yield a representative encyclopedic lexicon?
(e) How to achieve the right mix between encyclopedic and (collective) episodic knowledge?
(f) How to ‘dynamize’ the lexicon? The weight of the links between words is not frozen. During communication our focus shifts all the time. Likewise, any event (news) may have an impact on the evocation potential of a given word.
(g) Which lexical resources are relevant to support word finding? What layers of Levelt’s lexical access model are relevant, and to what extent do they need to be worked out? While we have a good understanding of the time course of lexical access, we still do lack details concerning the components at the various levels. For example, what are the specificities at the conceptual level? Which features typically get activated to evoke concepts like ‘justice’, ‘nationality’, ‘panda’?
(h) How to combine existing lexical resources to allow for their joined usage?
In sum, both theoretical/conceptual and empirical contributions are welcome.