allenai/alexafsm allenai/alexafsm

  • Stars
    star
    108
  • Rank 309,922 (Top 7 %)
  • Language
    Python
  • License
    Apache License 2.0
  • Created about 7 years ago
  • Updated about 1 year ago
Twitter logo Share LinkedIn logo Share Facebook logo Share
allenai/alexafsm
github

allenai

Reviews

There are no reviews yet. Be the first to send feedback to the community and the maintainers!

Repository Details

With alexafsm, developers can model dialog agents with first-class concepts such as states, attributes, transition, and actions. alexafsm also provides visualization and other tools to help understand, test, debug, and maintain complex FSM conversations.

alexafsm

  • Finite-state machine library for building complex Alexa conversations.
  • Free software: Apache Software License 2.0.

Dialog agents need to keep track of the various pieces of information to make decisions how to respond to a given user input. This is referred to as context, session, or state tracking. As the dialog complexity increases, this state-tracking logic becomes harder to write, debug, and maintain. This library takes the finite-state machine design approach to address this complexity. Developers using this library can model dialog agents with first-class concepts such as states, attributes, transition, and actions. Visualization and other tools are also provided to help understand and debug complex FSM conversations.

Also check out our blog post.

Features

  • FSM-based library for building Alexa skills with complex dialog state tracking.
  • Tools to validate, visualize, and print the FSM graph.
  • Support analytics with VoiceLabs.
  • Can be paired with any Python server library (Flask, CherryPy, etc.)
  • Written in Python 3.6 (primarily for type annotation and string interpolation).

Getting Started

Install from PyPi:

pip install alexafsm

Consult the Alexa skill search skill in the tests directory for details of how to write an alexafsm skill. An Alexa skill is composed of the following three classes: SessionAttributes, States, and Policy.

SessionAttributes

SessionAttributes is a class that holds session attributes (alexa_request['session']['attributes']) and any information we need to keep track of dialog state.

  • The core attributes are intent, slots, and state.
  • intent and slots map directly to Alexa's concepts.
  • slots should be of type Slots, which in turn is defined as a named tuple, one field for each slot type. In the skill search example, Slots = namedtuple('Slots', ['query', 'nth']). This named tuple class should be specified in the class definition as slots_cls = Slots.
  • state holds the name of the current state in the state machine.
  • Each Alexa skill can contain arbitrary number of additional attributes. If an attribute is not meant to be sent back to Alexa server (e.g. so as to reduce the payload size), it should be added to not_sent_fields. In the skill search example, searched and first_time are not sent to Alexa server.

See the implementation of skill search skill's SessionAttributes

States

States is a class that specifies most of the FSM and its behavior. It holds a reference to a SessionAttributes object, the type of which is specified by overriding the session_attributes_cls class attribute. The FSM is specified by a list of parameter-less methods. Consider the following method:

@with_transitions(
    {
        'trigger': NEW_SEARCH,
        'source': '*',
        'prepare': 'm_search',
        'conditions': 'm_has_result_and_query'
    },
    {
        'trigger': NTH_SKILL,
        'source': '*',
        'conditions': 'm_has_nth',
        'after': 'm_set_nth'
    },
    {
        'trigger': PREVIOUS_SKILL,
        'source': '*',
        'conditions': 'm_has_previous',
        'after': 'm_set_previous'
    },
    {
        'trigger': NEXT_SKILL,
        'source': '*',
        'conditions': 'm_has_next',
        'after': 'm_set_next'
    },
    {
        'trigger': amazon_intent.NO,
        'source': 'has_result',
        'conditions': 'm_has_next',
        'after': 'm_set_next'
    }
)
def has_result(self) -> response.Response:
    """Offer a preview of a skill"""
    attributes = self.attributes
    query = attributes.query
    skill = attributes.skill
    asked_for_speech = ''
    if attributes.first_time_presenting_results:
        asked_for_speech = _you_asked_for(query)
    if attributes.number_of_hits == 1:
        skill_position_speech = 'The only skill I found is'
    else:
        skill_position_speech = f'The {ENGLISH_NUMBERS[attributes.skill_cursor]} skill is'
        if attributes.first_time_presenting_results:
            if attributes.number_of_hits > 6:
                num_hits = f'Here are the top {MAX_SKILLS} results.'
            else:
                num_hits = f'I found {len(attributes.skills)} skills.'
            skill_position_speech = f'{num_hits} {skill_position_speech}'
    return response.Response(
        speech=f"{asked_for_speech} "
               f" {skill_position_speech} {_get_verbal_skill(skill)}."
               f" {HEAR_MORE}",
        card=f"Search for {query}",
        card_content=f"""
        Top result: {skill.name}

        {_get_highlights(skill)}
        """,
        reprompt=DEFAULT_PROMPT
    )

Each method encodes the following:

  • The name of the method is also the name of a state (describing) in the FSM.
  • The method may be decorated with one or several transitions, using with_transitions decorators. Transitions can be inbound (source needs to be specified) or outbound (dest needs to be specified).
  • Each method returns a Response object which is sent to Alexa.
  • Transitions can be specified with prepare and conditions attributes. See https://github.com/tyarkoni/transitions for detailed documentations. The values of these attributes are parameter-less methods of the Policy class.
  • The prepare methods are responsible for "actions" of the FSM such as querying a database. The after methods are responsible for updating the state after the transition completes. They are the only methods responsible for side-effects, e.g. modifying the attributes of the states. This design facilitates ease of debugging.

Policy

Policy is the class that holds everything together. It contains a reference to a States object, the type of which is specified by overriding the states_cls class attribute. A Policy object initializes itself by constructing a FSM based on the States type. Policy class contains the following key methods:

  • handle takes an Alexa request, parses it, and hands over all intent requests to execute method.
  • execute updates the policy's internal state with the request's details (intent, slots, session attributes), then calls trigger to make the state transition. It then looks up the corresponding response generating methods of the States class to generate a response for Alexa.
  • initialize will initialize a policy without any request.
  • validate performs validation of a policy object based on Policy class definition and a intent schema json file. It looks for intents that are not handled, invalid source/dest/prepare specifications, and unreachable states. The test in test_skillsearch.py performs such validation as a test of alexafsm.

The Alexa skill search skill in the tests directory also contains a Flask-based server that shows how to use Policy in five lines of code:

@app.route('/', methods=['POST'])
def main():
    req = flask_request.json
    policy = Policy.initialize()
    return json.dumps(policy.handle(req, settings.vi)).encode('utf-8')

Other Tools

alexafsm supports validation, graph visualization, and printing of the FSM.

Validation

Simply initialize a Policy before calling validate. This function takes as input the path to the skill's Alexa intent schema json file and performs the following checks:

  • All Alexa intents have corresponding events/triggers in the FSM.
  • All states have either inbound or outbound transitions.
  • All transitions are specified with valid source and destination states.
  • All conditions and prepare actions are handled with methods in the Policy class.

Change Detection with Record and Playback

When making code changes that are not supposed to impact a skill's dialog logic, we may want a tool to check that the skill's logic indeed stay the same. This is done by first recording (SkillSettings().record = True) one or several sessions, making the code change, then checking if the changed code still produces the same set of dialogs (SkillSettings().playback = True). During playback, calls to databases such as ElasticSearch can be fulfilled from data read from files generated during the recording. This is done by decorating the database call with recordable function. See the ElasticSearch call in Skill Search for an example usage.

Graph Visualization

alexafsm uses the transitions library's API to draw the FSM graph. For example, the skill search skill's FSM can be visualized using the graph.py. invoked from graph.sh. The resulting graph is displayed follow:

FSM Example

Graph Printout

For complex graphs, it may be easier to inspect the FSM in text format. Use the print_machine method to accomplish this. The output for the skill search skill is below:

Machine states:
	bad_navigate, describe_ratings, describing, exiting, has_result, helping, initial, is_that_all, no_query_search, no_result, search_prompt

Events and transitions:

Event: NthSkill
	Source: bad_navigate
		bad_navigate -> bad_navigate, conditions: ['m_has_nth']
		bad_navigate -> has_result, conditions: ['m_has_nth']
	Source: describe_ratings
		describe_ratings -> bad_navigate, conditions: ['m_has_nth']
		describe_ratings -> has_result, conditions: ['m_has_nth']
	Source: describing
		describing -> bad_navigate, conditions: ['m_has_nth']
		describing -> has_result, conditions: ['m_has_nth']
	Source: exiting
		exiting -> bad_navigate, conditions: ['m_has_nth']
		exiting -> has_result, conditions: ['m_has_nth']
	Source: has_result
		has_result -> bad_navigate, conditions: ['m_has_nth']
		has_result -> has_result, conditions: ['m_has_nth']
	Source: helping
		helping -> bad_navigate, conditions: ['m_has_nth']
		helping -> has_result, conditions: ['m_has_nth']
	Source: initial
		initial -> bad_navigate, conditions: ['m_has_nth']
		initial -> has_result, conditions: ['m_has_nth']
	Source: is_that_all
		is_that_all -> bad_navigate, conditions: ['m_has_nth']
		is_that_all -> has_result, conditions: ['m_has_nth']
	Source: no_query_search
		no_query_search -> bad_navigate, conditions: ['m_has_nth']
		no_query_search -> has_result, conditions: ['m_has_nth']
	Source: no_result
		no_result -> bad_navigate, conditions: ['m_has_nth']
		no_result -> has_result, conditions: ['m_has_nth']
	Source: search_prompt
		search_prompt -> bad_navigate, conditions: ['m_has_nth']
		search_prompt -> has_result, conditions: ['m_has_nth']
Event: PreviousSkill
	Source: bad_navigate
		bad_navigate -> bad_navigate, conditions: ['m_has_previous']
		bad_navigate -> has_result, conditions: ['m_has_previous']
	Source: describe_ratings
		describe_ratings -> bad_navigate, conditions: ['m_has_previous']
		describe_ratings -> has_result, conditions: ['m_has_previous']
	Source: describing
		describing -> bad_navigate, conditions: ['m_has_previous']
		describing -> has_result, conditions: ['m_has_previous']
	Source: exiting
		exiting -> bad_navigate, conditions: ['m_has_previous']
		exiting -> has_result, conditions: ['m_has_previous']
	Source: has_result
		has_result -> bad_navigate, conditions: ['m_has_previous']
		has_result -> has_result, conditions: ['m_has_previous']
	Source: helping
		helping -> bad_navigate, conditions: ['m_has_previous']
		helping -> has_result, conditions: ['m_has_previous']
	Source: initial
		initial -> bad_navigate, conditions: ['m_has_previous']
		initial -> has_result, conditions: ['m_has_previous']
	Source: is_that_all
		is_that_all -> bad_navigate, conditions: ['m_has_previous']
		is_that_all -> has_result, conditions: ['m_has_previous']
	Source: no_query_search
		no_query_search -> bad_navigate, conditions: ['m_has_previous']
		no_query_search -> has_result, conditions: ['m_has_previous']
	Source: no_result
		no_result -> bad_navigate, conditions: ['m_has_previous']
		no_result -> has_result, conditions: ['m_has_previous']
	Source: search_prompt
		search_prompt -> bad_navigate, conditions: ['m_has_previous']
		search_prompt -> has_result, conditions: ['m_has_previous']
Event: NextSkill
	Source: bad_navigate
		bad_navigate -> bad_navigate, conditions: ['m_has_next']
		bad_navigate -> has_result, conditions: ['m_has_next']
	Source: describe_ratings
		describe_ratings -> bad_navigate, conditions: ['m_has_next']
		describe_ratings -> has_result, conditions: ['m_has_next']
	Source: describing
		describing -> bad_navigate, conditions: ['m_has_next']
		describing -> has_result, conditions: ['m_has_next']
	Source: exiting
		exiting -> bad_navigate, conditions: ['m_has_next']
		exiting -> has_result, conditions: ['m_has_next']
	Source: has_result
		has_result -> bad_navigate, conditions: ['m_has_next']
		has_result -> has_result, conditions: ['m_has_next']
	Source: helping
		helping -> bad_navigate, conditions: ['m_has_next']
		helping -> has_result, conditions: ['m_has_next']
	Source: initial
		initial -> bad_navigate, conditions: ['m_has_next']
		initial -> has_result, conditions: ['m_has_next']
	Source: is_that_all
		is_that_all -> bad_navigate, conditions: ['m_has_next']
		is_that_all -> has_result, conditions: ['m_has_next']
	Source: no_query_search
		no_query_search -> bad_navigate, conditions: ['m_has_next']
		no_query_search -> has_result, conditions: ['m_has_next']
	Source: no_result
		no_result -> bad_navigate, conditions: ['m_has_next']
		no_result -> has_result, conditions: ['m_has_next']
	Source: search_prompt
		search_prompt -> bad_navigate, conditions: ['m_has_next']
		search_prompt -> has_result, conditions: ['m_has_next']
Event: AMAZON.NoIntent
	Source: has_result
		has_result -> bad_navigate, conditions: ['m_has_next']
		has_result -> has_result, conditions: ['m_has_next']
	Source: describe_ratings
		describe_ratings -> is_that_all
	Source: describing
		describing -> search_prompt
	Source: is_that_all
		is_that_all -> search_prompt
Event: DescribeRatings
	Source: bad_navigate
		bad_navigate -> describe_ratings, conditions: ['m_has_result']
	Source: describe_ratings
		describe_ratings -> describe_ratings, conditions: ['m_has_result']
	Source: describing
		describing -> describe_ratings, conditions: ['m_has_result']
	Source: exiting
		exiting -> describe_ratings, conditions: ['m_has_result']
	Source: has_result
		has_result -> describe_ratings, conditions: ['m_has_result']
	Source: helping
		helping -> describe_ratings, conditions: ['m_has_result']
	Source: initial
		initial -> describe_ratings, conditions: ['m_has_result']
	Source: is_that_all
		is_that_all -> describe_ratings, conditions: ['m_has_result']
	Source: no_query_search
		no_query_search -> describe_ratings, conditions: ['m_has_result']
	Source: no_result
		no_result -> describe_ratings, conditions: ['m_has_result']
	Source: search_prompt
		search_prompt -> describe_ratings, conditions: ['m_has_result']
Event: AMAZON.YesIntent
	Source: has_result
		has_result -> describing
	Source: describe_ratings
		describe_ratings -> describing
	Source: describing
		describing -> exiting
	Source: is_that_all
		is_that_all -> exiting
Event: AMAZON.CancelIntent
	Source: no_result
		no_result -> exiting
	Source: search_prompt
		search_prompt -> exiting
	Source: is_that_all
		is_that_all -> exiting
	Source: bad_navigate
		bad_navigate -> exiting
	Source: no_query_search
		no_query_search -> exiting
	Source: describing
		describing -> is_that_all
	Source: has_result
		has_result -> is_that_all
	Source: describe_ratings
		describe_ratings -> is_that_all
	Source: initial
		initial -> search_prompt
	Source: helping
		helping -> search_prompt
Event: AMAZON.StopIntent
	Source: no_result
		no_result -> exiting
	Source: search_prompt
		search_prompt -> exiting
	Source: is_that_all
		is_that_all -> exiting
	Source: bad_navigate
		bad_navigate -> exiting
	Source: no_query_search
		no_query_search -> exiting
	Source: describing
		describing -> is_that_all
	Source: has_result
		has_result -> is_that_all
	Source: describe_ratings
		describe_ratings -> is_that_all
	Source: initial
		initial -> search_prompt
	Source: helping
		helping -> search_prompt
Event: NewSearch
	Source: bad_navigate
		bad_navigate -> exiting, conditions: ['m_searching_for_exit']
		bad_navigate -> has_result, prepare: ['m_search'], conditions: ['m_has_result_and_query']
		bad_navigate -> no_query_search, conditions: ['m_no_query_search']
		bad_navigate -> no_result, prepare: ['m_search'], conditions: ['m_no_result']
	Source: describe_ratings
		describe_ratings -> exiting, conditions: ['m_searching_for_exit']
		describe_ratings -> has_result, prepare: ['m_search'], conditions: ['m_has_result_and_query']
		describe_ratings -> no_query_search, conditions: ['m_no_query_search']
		describe_ratings -> no_result, prepare: ['m_search'], conditions: ['m_no_result']
	Source: describing
		describing -> exiting, conditions: ['m_searching_for_exit']
		describing -> has_result, prepare: ['m_search'], conditions: ['m_has_result_and_query']
		describing -> no_query_search, conditions: ['m_no_query_search']
		describing -> no_result, prepare: ['m_search'], conditions: ['m_no_result']
	Source: exiting
		exiting -> exiting, conditions: ['m_searching_for_exit']
		exiting -> has_result, prepare: ['m_search'], conditions: ['m_has_result_and_query']
		exiting -> no_query_search, conditions: ['m_no_query_search']
		exiting -> no_result, prepare: ['m_search'], conditions: ['m_no_result']
	Source: has_result
		has_result -> exiting, conditions: ['m_searching_for_exit']
		has_result -> has_result, prepare: ['m_search'], conditions: ['m_has_result_and_query']
		has_result -> no_query_search, conditions: ['m_no_query_search']
		has_result -> no_result, prepare: ['m_search'], conditions: ['m_no_result']
	Source: helping
		helping -> exiting, conditions: ['m_searching_for_exit']
		helping -> has_result, prepare: ['m_search'], conditions: ['m_has_result_and_query']
		helping -> no_query_search, conditions: ['m_no_query_search']
		helping -> no_result, prepare: ['m_search'], conditions: ['m_no_result']
	Source: initial
		initial -> exiting, conditions: ['m_searching_for_exit']
		initial -> has_result, prepare: ['m_search'], conditions: ['m_has_result_and_query']
		initial -> no_query_search, conditions: ['m_no_query_search']
		initial -> no_result, prepare: ['m_search'], conditions: ['m_no_result']
	Source: is_that_all
		is_that_all -> exiting, conditions: ['m_searching_for_exit']
		is_that_all -> has_result, prepare: ['m_search'], conditions: ['m_has_result_and_query']
		is_that_all -> no_query_search, conditions: ['m_no_query_search']
		is_that_all -> no_result, prepare: ['m_search'], conditions: ['m_no_result']
	Source: no_query_search
		no_query_search -> exiting, conditions: ['m_searching_for_exit']
		no_query_search -> has_result, prepare: ['m_search'], conditions: ['m_has_result_and_query']
		no_query_search -> no_query_search, conditions: ['m_no_query_search']
		no_query_search -> no_result, prepare: ['m_search'], conditions: ['m_no_result']
	Source: no_result
		no_result -> exiting, conditions: ['m_searching_for_exit']
		no_result -> has_result, prepare: ['m_search'], conditions: ['m_has_result_and_query']
		no_result -> no_query_search, conditions: ['m_no_query_search']
		no_result -> no_result, prepare: ['m_search'], conditions: ['m_no_result']
	Source: search_prompt
		search_prompt -> exiting, conditions: ['m_searching_for_exit']
		search_prompt -> has_result, prepare: ['m_search'], conditions: ['m_has_result_and_query']
		search_prompt -> no_query_search, conditions: ['m_no_query_search']
		search_prompt -> no_result, prepare: ['m_search'], conditions: ['m_no_result']
Event: AMAZON.HelpIntent
	Source: bad_navigate
		bad_navigate -> helping
	Source: describe_ratings
		describe_ratings -> helping
	Source: describing
		describing -> helping
	Source: exiting
		exiting -> helping
	Source: has_result
		has_result -> helping
	Source: helping
		helping -> helping
	Source: initial
		initial -> helping
	Source: is_that_all
		is_that_all -> helping
	Source: no_query_search
		no_query_search -> helping
	Source: no_result
		no_result -> helping
	Source: search_prompt
		search_prompt -> helping

More Repositories

1

allennlp

An open-source NLP research library, built on PyTorch.
Python
11,691
star
2

OLMo

Modeling, training, eval, and inference code for OLMo
Python
3,949
star
3

RL4LMs

A modular RL library to fine-tune language models to human preferences
Python
2,020
star
4

longformer

Longformer: The Long-Document Transformer
Python
1,955
star
5

bilm-tf

Tensorflow implementation of contextualized word representations from bi-directional language models
Python
1,621
star
6

scispacy

A full spaCy pipeline and models for scientific/biomedical documents.
Python
1,566
star
7

bi-att-flow

Bi-directional Attention Flow (BiDAF) network is a multi-stage hierarchical process that represents context at different levels of granularity and uses a bi-directional attention flow mechanism to achieve a query-aware context representation without early summarization.
Python
1,524
star
8

scibert

A BERT model for scientific text.
Python
1,432
star
9

ai2thor

An open-source platform for Visual AI.
C#
1,010
star
10

open-instruct

Python
932
star
11

XNOR-Net

ImageNet classification using binary Convolutional Neural Networks
Lua
839
star
12

mmc4

MultimodalC4 is a multimodal extension of c4 that interleaves millions of images with text.
Python
793
star
13

dolma

Data and tools for generating and inspecting OLMo pre-training data.
Python
774
star
14

s2orc

S2ORC: The Semantic Scholar Open Research Corpus: https://www.aclweb.org/anthology/2020.acl-main.447/
Python
745
star
15

scitldr

Python
734
star
16

natural-instructions

Expanding natural instructions
Python
690
star
17

visprog

Official code for VisProg (CVPR 2023 Best Paper!)
Python
642
star
18

papermage

library supporting NLP and CV research on scientific papers
Python
605
star
19

science-parse

Science Parse parses scientific papers (in PDF form) and returns them in structured form.
Java
566
star
20

writing-code-for-nlp-research-emnlp2018

A companion repository for the "Writing code for NLP Research" Tutorial at EMNLP 2018
Python
558
star
21

pdffigures2

Given a scholarly PDF, extract figures, tables, captions, and section titles.
Scala
514
star
22

allennlp-models

Officially supported AllenNLP models
Python
512
star
23

tango

Organize your experiments into discrete steps that can be cached and reused throughout the lifetime of your research project.
Python
507
star
24

specter

SPECTER: Document-level Representation Learning using Citation-informed Transformers
Python
495
star
25

objaverse-xl

🪐 Objaverse-XL is a Universe of 10M+ 3D Objects. Contains API Scripts for Downloading and Processing!
Python
490
star
26

dont-stop-pretraining

Code associated with the Don't Stop Pretraining ACL 2020 paper
Python
488
star
27

unified-io-2

Python
471
star
28

macaw

Multi-angle c(q)uestion answering
Python
451
star
29

document-qa

Python
420
star
30

scholarphi

An interactive PDF reader.
Python
410
star
31

deep_qa

A deep NLP library, based on Keras / tf, focused on question answering (but useful for other NLP too)
Python
405
star
32

acl2018-semantic-parsing-tutorial

Materials from the ACL 2018 tutorial on neural semantic parsing
402
star
33

unifiedqa

UnifiedQA: Crossing Format Boundaries With a Single QA System
Python
384
star
34

kb

KnowBert -- Knowledge Enhanced Contextual Word Representations
Python
359
star
35

pawls

Software that makes labeling PDFs easy.
Python
356
star
36

PeerRead

Data and code for Kang et al., NAACL 2018's paper titled "A Dataset of Peer Reviews (PeerRead): Collection, Insights and NLP Applications"
Python
354
star
37

naacl2021-longdoc-tutorial

Python
343
star
38

openie-standalone

Quality information extraction at web scale. Edit
Scala
329
star
39

python-package-template

A template repo for Python packages
Python
318
star
40

allenact

An open source framework for research in Embodied-AI from AI2.
Python
295
star
41

acl2022-zerofewshot-tutorial

293
star
42

ir_datasets

Provides a common interface to many IR ranking datasets.
Python
291
star
43

s2orc-doc2json

Parsers for scientific papers (PDF2JSON, TEX2JSON, JATS2JSON)
Python
290
star
44

beaker-cli

A collaborative platform for rapid and reproducible research.
Go
230
star
45

Holodeck

CVPR 2024: Language Guided Generation of 3D Embodied AI Environments.
Python
220
star
46

procthor

🏘️ Scaling Embodied AI by Procedurally Generating Interactive 3D Houses
Python
214
star
47

comet-atomic-2020

Python
212
star
48

FineGrainedRLHF

Python
209
star
49

fm-cheatsheet

Website for hosting the Open Foundation Models Cheat Sheet.
Python
207
star
50

spv2

Science-parse version 2
Python
206
star
51

scifact

Data and models for the SciFact verification task.
Python
206
star
52

OLMo-Eval

Evaluation suite for LLMs
Python
200
star
53

unified-io-inference

Jupyter Notebook
196
star
54

allennlp-demo

Code for the AllenNLP demo.
TypeScript
191
star
55

lumos

Code and data for "Lumos: Learning Agents with Unified Data, Modular Design, and Open-Source LLMs"
Python
190
star
56

citeomatic

A citation recommendation system that allows users to find relevant citations for their paper drafts. The tool is backed by Semantic Scholar's OpenCorpus dataset.
Jupyter Notebook
182
star
57

cartography

Dataset Cartography: Mapping and Diagnosing Datasets with Training Dynamics
Jupyter Notebook
180
star
58

savn

Learning to Learn how to Learn: Self-Adaptive Visual Navigation using Meta-Learning (https://arxiv.org/abs/1812.00971)
Python
175
star
59

vampire

Variational Methods for Pretraining in Resource-limited Environments
Python
173
star
60

objaverse-rendering

📷 Scripts for rendering Objaverse
Python
169
star
61

hidden-networks

Python
164
star
62

ScienceWorld

ScienceWorld is a text-based virtual environment centered around accomplishing tasks from the standardized elementary science curriculum.
Scala
156
star
63

vila

Incorporating VIsual LAyout Structures for Scientific Text Classification
Python
155
star
64

mmda

multimodal document analysis
Jupyter Notebook
154
star
65

cord19

Get started with CORD-19
149
star
66

PRIMER

The official code for PRIMERA: Pyramid-based Masked Sentence Pre-training for Multi-document Summarization
Python
145
star
67

dnw

Discovering Neural Wirings (https://arxiv.org/abs/1906.00586)
Python
139
star
68

tpu_pretrain

LM Pretraining with PyTorch/TPU
Python
129
star
69

deepfigures-open

Companion code to the paper "Extracting Scientific Figures with Distantly Supervised Neural Networks" 🤖
Python
129
star
70

catwalk

This project studies the performance and robustness of language models and task-adaptation methods.
Python
129
star
71

allentune

Hyperparameter Search for AllenNLP
Python
128
star
72

lm-explorer

interactive explorer for language models
Python
127
star
73

pdffigures

Command line tool to extract figures, tables, and captions from scholarly documents in PDF form.
C++
125
star
74

SciREX

Data/Code Repository for https://api.semanticscholar.org/CorpusID:218470122
Python
125
star
75

s2-folks

Public space for the user community of Semantic Scholar APIs to share scripts, report issues, and make suggestions.
125
star
76

scidocs

Dataset accompanying the SPECTER model
Python
124
star
77

gooaq

Question-answers, collected from Google
Python
116
star
78

OpenBookQA

Code for experiments on OpenBookQA from the EMNLP 2018 paper "Can a Suit of Armor Conduct Electricity? A New Dataset for Open Book Question Answering"
Python
113
star
79

allennlp-as-a-library-example

A simple example for how to build your own model using AllenNLP as a dependency.
Python
113
star
80

allennlp-semparse

A framework for building semantic parsers (including neural module networks) with AllenNLP, built by the authors of AllenNLP
Python
107
star
81

scicite

Repository for NAACL 2019 paper on Citation Intent prediction
Python
106
star
82

peS2o

Pretraining Efficiently on S2ORC!
105
star
83

multimodalqa

Python
102
star
84

commonsense-kg-completion

Python
102
star
85

real-toxicity-prompts

Jupyter Notebook
101
star
86

ai2thor-rearrangement

🔀 Visual Room Rearrangement
Python
97
star
87

embodied-clip

Official codebase for EmbCLIP
Python
97
star
88

aristo-mini

Aristo mini is a light-weight question answering system that can quickly evaluate Aristo science questions with an evaluation web server and the provided baseline solvers.
Python
96
star
89

s2search

The Semantic Scholar Search Reranker
Python
93
star
90

elastic

Python
91
star
91

reward-bench

RewardBench: the first evaluation tool for reward models.
Python
90
star
92

flex

Few-shot NLP benchmark for unified, rigorous eval
Python
89
star
93

gpv-1

A task-agnostic vision-language architecture as a step towards General Purpose Vision
Jupyter Notebook
89
star
94

manipulathor

ManipulaTHOR, a framework that facilitates visual manipulation of objects using a robotic arm
Jupyter Notebook
86
star
95

medicat

Dataset of medical images, captions, subfigure-subcaption annotations, and inline textual references
Python
85
star
96

propara

ProPara (Process Paragraph Comprehension) dataset and models
Python
82
star
97

allennlp-guide

Code and material for the AllenNLP Guide
Python
81
star
98

hierplane

A tool for visualizing trees, tailored specifically to the analysis of parse trees.
JavaScript
81
star
99

S2AND

Semantic Scholar's Author Disambiguation Algorithm & Evaluation Suite
Python
78
star
100

ARC-Solvers

ARC Question Solvers
Python
78
star