Jonny: Undo revision 478 by Jonny (talk)

2022-10-26T21:13:55Z

Undo revision 478 by Jonny (talk)

← Older revision		Revision as of 21:13, 26 October 2022
Line 1:		Line 1:
	{{Participant		{{Participant
	\|Timezone=America/~~Chicago~~ (~~GMT−06~~:00/~~GMT−05~~:00)		\|Timezone=America/Los Angeles (GMT−08:00/GMT−07:00)
	\|Affiliation=Viva Translate		\|Affiliation=Viva Translate
	~~\|Projects=SoftCite~~
	\|Table Assignment=Table 3		\|Table Assignment=Table 3
	}}		}}

Jonny at 20:50, 26 October 2022

2022-10-26T20:50:24Z

← Older revision		Revision as of 20:50, 26 October 2022
Line 1:		Line 1:
	{{Participant		{{Participant
	\|Timezone=America/~~Los Angeles~~ (~~GMT−08~~:00/~~GMT−07~~:00)		\|Timezone=America/Chicago (GMT−06:00/GMT−05:00)
	\|Affiliation=Viva Translate		\|Affiliation=Viva Translate
			\|Projects=SoftCite
			\|Table Assignment=Table 3
	}}		}}
	{{Workshop Submission		{{Workshop Submission

Jonny at 06:37, 21 October 2022

2022-10-21T06:37:36Z

← Older revision		Revision as of 06:37, 21 October 2022
Line 29:		Line 29:

	A demo of the extension is in the Github repo.		A demo of the extension is in the Github repo.
	\|Organizer Topics=Discovery, Accessibility, Transdisciplinarity, NLP, Machine Learning, Review		\|Organizer Topics=Discovery, Accessibility, Transdisciplinarity, NLP, Machine Learning, Review, Scholcomm
	}}		}}

Jonny at 06:34, 21 October 2022

2022-10-21T06:34:07Z

← Older revision		Revision as of 06:34, 21 October 2022
Line 29:		Line 29:

	A demo of the extension is in the Github repo.		A demo of the extension is in the Github repo.
			\|Organizer Topics=Discovery, Accessibility, Transdisciplinarity, NLP, Machine Learning, Review
	}}		}}

Jonny: Created page with "{{Participant |Timezone=America/Los Angeles (GMT−08:00/GMT−07:00) |Affiliation=Viva Translate }} {{Workshop Submission |Interest=High-context textual information increases the barrier to comprehension for the majority of people to understand a concept. In particular, academic papers have such a high density of context per word that it is difficult for newcomers to learn about a field. This is at odds with how humans communicate in a live conversation. Typically, a..."

2022-10-20T04:45:56Z

Created page with "{{Participant |Timezone=America/Los Angeles (GMT−08:00/GMT−07:00) |Affiliation=Viva Translate }} {{Workshop Submission |Interest=High-context textual information increases the barrier to comprehension for the majority of people to understand a concept. In particular, academic papers have such a high density of context per word that it is difficult for newcomers to learn about a field. This is at odds with how humans communicate in a live conversation. Typically, a..."

New page

{{Participant
|Timezone=America/Los Angeles (GMT−08:00/GMT−07:00)
|Affiliation=Viva Translate
}}
{{Workshop Submission
|Interest=High-context textual information increases the barrier to comprehension for the majority of people to understand a concept. In particular, academic papers have such a high density of context per word that it is difficult for newcomers to learn about a field.

This is at odds with how humans communicate in a live conversation. Typically, a layperson can understand the general concepts and purpose of a paper by speaking with the paper's author for 5-10 minutes. However, if they were only looking at the paper, this process would be daunting and take much longer. The learner would need to look up jargon, click through references, and continuously re-read complex explanations in order to understand the author's intent.

This extra friction intimidates newcomers away from learning, limits the public from engaging in scientific discussion, and discourages cross-discipline collaboration. Reducing the cognitive barrier to entry would significantly improve scientific progress. Decreasing the upfront cost to learn about a new field would lead to greater cross-pollination of fields, more scrutiny on the ideas presented, and more people participating in these research fields.

This workshop will be structured around sharing the tool we've built using large language models (LLMs) to simplify academic jargon into everyday conversational words. Other participants will be able to give feedback on the tool, share their thoughts on how we might use LLMs to lower the barrier to entry in scientific fields, and potentially partner together with those who are working on similar ideas.
|Frame=Tool-builder
|Materials=https://github.com/epli-org/epli

Science has produced incredible innovations, yet there is incredible amounts of progress to be made on the scientific process itself. Specifically, scientific knowledge is still spread via academic papers, and people still read them as if they are textual documents that live in your computer, as opposed to interactive elements for querying and collaboration. We hope to change that by building a Chrome extension that uses large language models to make it easier to access the relevant information from a paper.

When looking at papers, researchers often find themselves needing to manually look up the referenced papers in separate tabs and search for the relevant section within them, as the links directly to the papers aren’t usually provided by default. This causes a lot of context switching, and generally increases friction to learn new concepts. To alleviate this, our Chrome extension allows a user to hover over a reference within a paper, and presents a pop-up containing:

- The direct link to the paper
- The abstract of the paper
- A context-sensitive GPT-3 generated summary of the referenced paper

This allows the reader to quickly understand the relevant concepts within the cited resource, so that they can immediately resume their processing of the current paper. This also reduces the number of clicks steps they will need to go through in order to access that paper if they so desire.

Scientific papers also contain "jargon", and this often forces readers to spend more time than necessary getting up to speed with the language of a paper before they can actually understand the concepts presented within. When our users hover over scientific jargon within a paper, a pop-up will be presented that shows the definition of the piece of jargon, along with links to locations where it’s defined previously. This is similar to jump-to-definition in IDEs, except applied for concepts in a paper.

In the long term, we plan on enabling a commenting system on top of papers, so that users can ask and answer questions on top of papers, while earning reputation for doing so. The features outlined above will be useful for getting initial interest in the product.

A demo of the extension is in the Github repo.
}}

Belinda Mo - Revision history

Jonny: Undo revision 478 by Jonny (talk)

Jonny at 20:50, 26 October 2022

Jonny at 06:37, 21 October 2022

Jonny at 06:34, 21 October 2022