About this Research Topic
Language and Computation invites submissions of position papers on significant, controversial, and far-reaching issues related to the study of language and linguistics using computational models and methods.
The purpose of position papers is to stimulate discussion and debate and to attract attention to new ideas and perspectives. In this spirit, the position papers we seek should offer novel and fresh viewpoints on the field, open up new research questions, compare and critique existing scientific paradigms, or express reasoned opinions about the methodological issues in language and computation research. Cross-disciplinary papers are particularly sought.
To encourage debate and discussion in the community, the editors will solicit commentaries and reactions to accepted position papers and responses from the authors.
The importance of such open and vigorous discussion of the state of scholarship and research into language and computation cannot be overstated. It is becoming widely recognized that while computation has formed a powerful paradigm for research into language, the state of our field is far from optimal. Specialist research communities in our field are largely disjoint, each supported by its own incentive mechanisms, methodological assumptions, institutional structures, and discourse norms. Moreover, the results in many of these communities are rapidly approaching local optima. The following trends are much too familiar:
• Optimizing research to particular metrics and datasets, losing sight of the underlying problem and measuring progress by percentage increase in accuracy on task-oriented benchmarks rather than advancing understanding of phenomena or needs of real world applications.
• Working on solving scoped-down versions of linguistic issues, and then treating the redefinitions as the main problem. For example, in many quarters, coreference resolution is treated as the question of how to determine a pronoun’s antecedent, though that is but one of many coreference phenomena.
• Preferences for “form over content” – focusing on metalanguages of language description rather than language material as such; this norm enables much work in theoretical linguistics and philosophy of language to avoid addressing verifiable empirical content.
• Treating language as just an application domain for machine learning or other computational analysis techniques, with no consideration of what linguists already know about language.
• Looking for correlational patterns in massive datasets with little-to-no theoretical framework and no emphasis on explanation.
• Putting computational efficiency before descriptive and explanatory adequacy, thus incentivizing a focus on narrow problems.
We call for position papers (and responses to them) that will influence thought, strategies, and norms in language research by articulating how to incentivize and facilitate long-term progress of the field and to escape trends such as those noted above that can hinder this objective. Potential topic areas include, but are certainly not limited to:
• Properly defining the scope of phenomena to be studied, negotiating tradeoffs between feasibility, depth of understanding, and generality;
• Ensuring that methodology is empirically grounded but doesn’t stifle new ideas;
• Balancing short-term verifiable research progress with long-term research programs;
• The relationship of theory development, empirical validation, and application development and testing;
• How best to develop, validate, balance between, and/or integrate micro-theoretical treatments of particular phenomena and general theoretical frameworks;
• How (and whether) to deal with the question of typological validity – ensuring that theories, analytical methods, and empirical results are generalizable across different languages and families.
To submit, authors can click the "Submit your manuscript" link at the top of this page and select the relevant Article Type from the list (e.g. Reviews, Perspective Papers, Brief Research Reports, General Commentaries, Opinion, etc. See the full list of Article Types here: https://www.frontiersin.org/journals/big-data#article-types). All papers are fully peer reviewed.
Keywords: nlp, natural language processing, language and computation
Important Note: All contributions to this Research Topic must be within the scope of the section and journal to which they are submitted, as defined in their mission statements. Frontiers reserves the right to guide an out-of-scope manuscript to a more suitable section or journal at any stage of peer review.