Paraphrasing with GPT-3

Prompting
1

I’ve been trying to get GPT-3 to paraphrase passages of text, with no success.

With all prompts i tried, at low temperatures, GPT-3 repeats the input text. At high temperatures, it starts to produce factually incorrect sentences. For the use case I have in mind, the model needs to produce factually correct output in at least 95% of the time. But the text should also be sufficiently different from the original.

Example prompt
Paraphrase the original passage.
""""""
Original passage:

""""""
Paraphrased passage:

Has anyone cracked paraphrasing? I also tried one-shot and few-shot prompts but with similar results.

2 Likes
2

I’ve had great success with CURIE-INSTRUCT. The prompt is simply “Generate a very detailed summary of the following:” and then a few examples.

2 Likes
4

That’s an important note to keep in mind, @m-a.schenk!

@daveshapautomator that’s a great prompt. I’ve had a lot of success using davinci-instruct-beta with no examples at all (with penalties to the max to avoid repeating the prompt, and a medium temperature).

Screenshot 2021-06-05 at 10.30.34
Screenshot 2021-06-05 at 10.30.342418×436 48.8 KB

5 Likes
5

did you use stop words, start words? I mean, what would be the details of the parameters Joey?

6

Just the default parameters, besides what I described.

1 Like
7

Thanks joey! My prompts returns similar results.

The issue is, that as soon as the model “coincidentally” returns one sentence, that is the same as in the input, all the remaining sentences are exact copies (which seems natural given the “predict next word” objective of the model). See e.g. the last two sentences in your output, they match the input exactly.

8

Glad to help!

This is just a starting point, and I’d definitely add high-quality, varied examples to help the API.

Edit: To add to that, you could also re-phrase one sentence at a time, circumventing the problem of entire paragraphs not being properly re-phrased. That said, this is entirely different from the earlier suggestion of providing examples, so I’d try both out, depending on your exact needs.

1 Like
9

Informal: Hulu has a big user base.

Original: Hulu commands a sizeable market.

Rewrite: (Keywords: wields, reach) Hulu wields a vast reach.

Informal: Milk is good for your bones.

Original: Milk strengthens one’s bones.

Rewrite: (Keywords: documented, bolster) Milk has been shown to bolster one’s bone strength.

Informal: GMOs are seen as bad, but that’s not accurate.

Original: GMOs command a negative reputation, but this sentiment is inaccurate.

Rewrite: (Keywords: tainted, inaccurate) GMOs bear a tainted legacy, but this assessment is unfounded.

4 Likes
10

This is a pretty good prompt! :+1:

11

If you’re paraphrasing a list of items, GPT-3 won’t shorten the source material. A list can be anything that has similar meaning.

For example: “I went to the foyer. Then I put on my shoes. Then I opened the door. Then I…”

Summarization works best when each sentence adds to the overall theme, without adding too much new repeatable information. Think narrative arc.

For example: “He had a tattoo on his left wrist. It was a star from his neighborhood childhood, he told me. He seemed downcast when he looked down at it, as if he was appraising a faded picture of long dead relatives in some attic. I’ve always remembered that tattoo…”

3 Likes
12

Is there any reason why the Content rephraser is capped at only 30 tokens?

A lot of times, our input sentence easily exceed that - and it’s precisely why we need the AI’s help to rephrase and break down the sentence.

It’s quite unusable at 30 tokens.

1 Like
14

Works for me using text-dacinci-003 temperature 0.7 and the prompt:

paraphrase “Your text here”

1 Like
15

Hi All. I have a question regarding paraphrasing. English is not my native language. I ask AI help me check grammar or paraphrase my original sentences. In some cases, it doesn’t change, in some provides some suggestions. I use some of these suggestions in writing my book. I try not to copy-paste but modify slightly. Are there any ethical issues?

16

I’ve created a simple tool which does the same, I used turbo with simple prompts with examples. https://revrite.ai/

1 Like
18

Theoretical Lens: Socio-Cognitive Perspective

The socio-cognitive perspective provides a theoretical lens to examine the effects of online games on the academic competencies of grade 9 students. This perspective emphasizes the interplay between social factors and cognitive processes in shaping individual behavior and learning outcomes. It considers how online games can influence students’ academic competencies through their interactions with peers, cognitive development, and information processing.

Peer Interaction and Collaborative Learning:
Online games often involve multiplayer features that encourage students to collaborate and communicate with their peers. Through these interactions, students can develop important social and cognitive skills, such as teamwork, problem-solving, and effective communication. These skills can transfer to academic settings and enhance their academic competencies, particularly in subjects that require collaboration and critical thinking.

Cognitive Development and Problem-Solving:
Many online games require players to solve complex problems, make strategic decisions, and adapt to changing situations. Engaging in such games can stimulate cognitive processes, including memory, attention, and executive functions. These cognitive skills can be beneficial for academic competencies, as they enhance students’ ability to analyze information, think critically, and solve academic problems effectively.

Information Processing and Digital Literacy:
Online games often involve navigating virtual environments, deciphering information, and making decisions based on available cues. These activities can improve students’ information processing skills and digital literacy, which are increasingly important in the digital age. Effective information processing and digital literacy can contribute to students’ ability to locate, evaluate, and use information from various sources, thus enhancing their academic competencies across different subjects.

Motivation and Engagement:
Online games are often designed to be highly engaging and rewarding, providing immediate feedback and a sense of achievement. This can enhance students’ motivation to learn and persist in challenging tasks. When students find academic subjects engaging and rewarding, they are more likely to invest effort, time, and attention, leading to improved academic competencies.

Balancing Online Game Use and Academic Demands:
While online games can have positive effects on academic competencies, it is crucial to address the potential negative impacts of excessive game use. Excessive gaming can lead to reduced time spent on academic tasks, sleep disturbances, and neglect of other important activities. Establishing a healthy balance between online game use and academic demands is essential to ensure that students derive benefits from gaming without compromising their academic competencies.

Overall, the socio-cognitive perspective highlights the potential positive impacts of online games on grade 9 students’ academic competencies. By fostering peer interaction, cognitive development, information processing, motivation, and engagement, online games can contribute to students’ overall academic success. However, it is important to consider the balance between game use and academic responsibilities to maximize the benefits while mitigating potential drawbacks.

19

The data that you gathered will be reliable and valid if the source is written by a trustworthy author and/or institution and if a test produces accurate results

1 Like
20

I have been trying to create a paraphrasing tool using text davinci.
Well the output responses are good but I dont want to alter the format content but text davincie does that.
I tried few prompts where i mentioned not change the format but its not working for me.

21

I would try GPT-3.5-turbo or GPT-4 for that kind of work, you can “instruct” them to keep the same format as they are instruct models.

22

The Language Acquisition Device (LAD), proposed by Noam Chomsky, is a theoretical concept suggesting an innate mental structure that enables humans to acquire language. Key features include universality, rapidity, and a critical period for language development.