Reading and Reviewing Papers › Module 1: Finding papers

Searching well

Lesson 2 of 3 · 2 min

Previous

Searching for papers is a skill. A few habits make it much faster.

Start with keywords, then grow the list

Write your question, then pull out the key ideas. For each one, list other words people use for it.

low-resource
Other words people use: low-resource, under-resourced, data-scarce
speech recognition
Other words people use: speech recognition, ASR, speech-to-text
Indian languages
Other words people use: Indic, Hindi, Tamil, Bengali, code-switched

When you find a good paper, look at the words in its title and abstract. Researchers in a field often use a fixed term you did not know, and that term opens up many more papers.

Search operators

Most search tools, including Google Scholar, understand some operators:

  • "exact phrase": put a phrase in quotes. "data leakage"
  • OR: either word. ASR OR "speech recognition"
  • minus: leave a word out. transformer -electrical
  • author: in Google Scholar, search by author. author:"Y Bengio"

You can also limit Google Scholar results to a range of years. For a fast-moving field like AI, start with the last three to five years, then go back for the classic papers.

Read the results smartly

  • Sort by relevance first, then look at recent work.
  • Read titles, then abstracts. Open the full paper only if the abstract fits.
  • High citation counts suggest influence, not correctness. New papers have few citations because they are new.

Save as you go

Every time you find a useful paper, save it straight away into your reference manager (module 4). Searching again for a paper you saw last week wastes a lot of time.

RememberKeep a short search diary: which tool, which words, how many results, how many useful. It helps you write the method of a literature review later.