INLP Workshop

Introduction

While general conversational intelligence (GCI) can be considered one of the core aspects of AGI, the fields of AGI and NLP currently have little overlap, with few existing AGI architectures capable of comprehending natural language and nearly all NLP systems founded upon specialized, hardcoded rules and language-specific frameworks. This workshop is centered around the idea of INLP, an extension of the interpretable AI (IAI) concept to NLP; INLP allows for acquisition of natural language, comprehension of textual communications, and production of textual messages in a reasonable and transparent way. The proposed presentations regarding Link Grammar (LG), unsupervised LG learning, interpretable NLG/NLS, and sentiment mining/topic matching cover various INLP methods that may bring a greater degree of GCI to proto-AGI pipelines.

Particular topics of interest include, but are not limited to:

Interpretability in dialogue systems
Intellectual probing methods for NLP
Interpretability in representation of language models
New metrics and Evaluation in NLP
New tools for interpretable benchmarking

Committee

Organizers

Anton Kolonin, Ph.D., AI and Blockchain Architect @ SingularityNET Foundation, Founder @ Aigents, Senior Lecturer and Lead Specialist @ Novosibirsk State University
Linas Vepstas, Ph.D., Research Scientist @ OpenCog Foundation
Tatiana Shavrina, AGI NLP Team Lead @ SberDevices, NLP Team Lead @ AIRI
Vignav Ramesh, NLP Research & Development Intern @ SingularityNET Foundation, Contributor @ Aigents

Tentative Schedule

The workshop will be structured as a hybrid event – both virtually and physically with in-person attendance at the Seattle (Washington, USA) conference venue, depending on COVID-19 regulations at the time of the conference. Each presentation is expected to be structured as a talk, between 30 minutes and one hour (depending on the number of accepted speakers) in duration, including Q&A time at the end. The overall schedule will be aligned to that of the AGI-22 conference.

ALL PARTICIPANTS (virtual AND in-person) must register at https://forms.gle/7tr34DuvW8KfD3hz6. The attendance information for virtual/online participants will be sent to registrants (those who fill out the INLP registration form) via the emails included in registrants’ form responses. Physical attendance requires additional registration at http://agi-conf.org/2022/registration/. Updates on the workshop organization and any questions during the workshop are expected on Telegram channel.

The following presentations are scheduled (US Pacific times):

09:00-10:00 Linas Vepstas: "Experimental Results on Unsupervised Grammar Induction"
10:00-10:30 Nikolay Mikhailovsky: "On Unsupervised Learning of Link Grammar Based Language Models"
10:30-11:00 Anton Kolonin: "Unsupervised lexicon and punctuation discovery"
11:00-11:30 Discussion / Coffee break
11:30-12:00 Victor Nosko: "Fast experts tuning: a better domain adaptation method for transformers efficient tuning"
12:00-12:30 Mukul Vishvas: "Social Media Sentiment Analysis for Cryptocurrency Market Prediction"
12:30-13:30 Discussion / Lunch

The workshop has ended. Watch the recording below:

Watch Workshop Recording

Call for Papers

The paper submission period is now over.

We have an open call for speakers to submit papers/presentations. The workshop scope may include extended presentations based on papers accepted for the main AGI-2022 conference and papers not accepted for the main conference, as well as preprints on arXiv or other publications/talk proposals submitted in alternative formats. The call for presentations will be open till July 20, 2022 with author notification by July 30.

To submit a paper or presentation, email inlp22@googlegroups.com the following information:

Full Name (e.g., John Doe)

Coauthor Names, comma-separated (e.g., Jacob Smith, Mary Anne)

Email (e.g., jdoe@google.com)

Presentation Title (e.g., Nerfies: Deformable Neural Radiance Fields)

Abstract (e.g., We present the first method capable of photorealistically reconstructing a non-rigidly deforming scene using photos/videos captured casually from mobile phones...)

Required time (e.g., 15-45 mins)

Presentation Slides and/or Full Paper (PDF Format)

Publications

Vepstas, L., Goertzel, B.: Learning Language from a Large (Unannotated) Corpus. arXiv:1401.3372 [cs.LG] (2014).
Vepstas, L.: Sheaves: A Topological Approach to Big Data. arXiv:1901.01341 [cs.LG] (2019).
Ramesh, V., Kolonin, A.: Natural Language Generation Using Link Grammar for General Conversational Intelligence. arXiv:2105.00830 [cs.CL] (2021).

Ramesh, V., Kolonin, A.: Interpretable Natural Language Segmentation Based on Link Grammar. In: 2020 Science and Artificial Intelligence conference (S.A.I.ence), pp. 25–32. IEEE, Novosibirsk (2020).

Sleator, D., Temperley, D.: Parsing English with a Link Grammar. In: Proceedings of the Third International Workshop on Parsing Technologies, pp. 277–292. Association for Computational Linguistics, Netherlands (1993).

Lian, R., et al.: Syntax-Semantic Mapping for General Intelligence: Language Comprehension as Hypergraph Homomorphism, Language Generation as Constraint Satisfaction. In: Bach, J., Goertzel, B., Iklé, M. (eds.) ARTIFICIAL GENERAL INTELLIGENCE 2012, LNCS, vol 7716, pp. 158–167. Springer, Heidelberg (2012).

Glushchenko, A., et al.: Unsupervised Language Learning in OpenCog. In: Iklé M., Franz, A., Rzepka, R., Goertzel, B. (eds.) ARTIFICIAL GENERAL INTELLIGENCE 2018, LNCS, vol 10999, pp. 109–118 Springer, Cham (2018).

Glushchenko, A., Suarez, A., Kolonin, A., Goertzel, B., Baskov, O.: Programmatic Link Grammar Induction for Unsupervised Language Learning. In: Hammer, P., Agrawal, P., Goertzel, B., Iklé, M. (eds.) ARTIFICIAL GENERAL INTELLIGENCE 2019, LNCS, vol 11654, pp. 111-120. Springer, Cham (2019).

Vepstas, L.: Sheaves: A Topological Approach to Big Data. arXiv:1901.01341 [cs.LG] (2019).

Lian, R., Goertzel, B., Vepstas, L., Hanson, D., Zhou, C.: Symbol Grounding via Chaining of Morphisms. arXiv:1703.04368 [cs.AI] (2017).

Vepstas, L., Goertzel, B.: Learning Language from a Large (Unannotated) Corpus. arXiv:1401.3372 [cs.CL] (2014).

Goertzel, B., et al.: A General Intelligence Oriented Architecture for Embodied Natural Language Processing. In: Proceedings of the 3rd Conference on Artificial General Intelligence, pp. 68-73. Atlantis Press (2010).

Goertzel, B., Suarez, A., Yu, G.: Guiding Symbolic Natural Language Grammar Induction via Transformer-Based Sequence Probabilities. arXiv:2005.12533 [cs.CL] (2020).

Freeman, R.: Parsing using a grammar of word association vectors. arXiv:1403.2152 [cs.CL] (2014).