add overview of inputs and outputs to readme

README.md

@@ -12,6 +12,38 @@ and we can't guarantee no breaking changes either of library interface or databa
 notably, the database schema will have a significant change to allow multiple JSON files to be loaded into the same
 database file.
 
+## Contents
+
+- [Collecting Twitter Data](#collecting-twitter-data)
+- [Input and Output](#input-and-output)
+- [Installation](#installation)
+- [Usage](#usage)
+- [About tidy_tweet](#about-tidy_tweet)
+
+## Collecting Twitter data
+
+If you do not have a preferred Twitter collection tool already, we recommend [Twarc](https://github.com/DocNow/twarc/). 
+tidy_tweet is designed to work directly with Twarc output. Other collection methods may work with tidy_tweet as long
+as they output the API result from Twitter with minimal alteration (see [Input and Output](#input-and-output)), however 
+at this time we do not have the resources to support Twitter data outputs from tools other than Twarc. 
+
+## Input and Output
+
+### Input: Twitter results pages
+
+tidy_tweet takes as input a series of JSON/dict objects, each object of which is a page of Twitter API v2 search or 
+timeline results. Typically, this will be a JSON file such as those output by `twarc2 search`.
+
+JSON files with multiple pages of results are expected to be newline-delimited, with each line being a distinct results
+page object, and no commas between top-level objects.
+
+### Output: Sqlite database of tweets and metadata
+
+After processing your Twitter results pages with tidy_tweet (see [Usage](#usage)), you will have an 
+[SQLite](https://sqlite.org/index.html) database file at the location you specified.
+
+Database schema will be published here as soon as the initial schema is finalised.
+
 ## Installation
 
 tidy_tweet is a Python package and can be installed with pip.