How do I get an Application Key?
I am a content provider, how do I get Appinions to index my data?
I am a content provider, how do I get Appinions to stop indexing my data?
I keep getting an "Over Rate Limit" or "Over Queries Per Second" error?
We cap usage for our free API at 5000 requests per day and 2 requests per second. If you need additional usage please Contact us
Does your service work for other languages like Chinese, Spanish or Esperanto?
We currently only process English. We hope to get to other languages soon (but not Esperanto. Never).
How does your technology work? How do you identify Opinions, Opinion-holders, etc?
We apply advanced Natural Language Processing techniques that come out many years of research at Cornell in Opinion Extraction and Sentiment Analysis. If you're interested in digging in, here are a few papers.
I am not getting back many results for my query. Why is that?
We currently index a variety of news, blogs and social media data using content aggregators. If you see gaps in our coverage let us know.
How come I find plenty of current results, but not much historial data?
We are focused on providing timely opinions on the "fresh web". Our API accesses opinions over the previous three months. Additionally, we are constantly attempting to increase our capacity and process new sources. That means we will often have more opinions for today than we had yesterday or last month.
How do I get back cleaner results with the Opinion Search API? How do I remove some of the noise?
Opinion-Extraction is not perfect, but we try to present all our results to you. For some applications it is helpful to tune the results towards precision by filtering out results according to more strict constraints. Some strategies that we've found useful when building appications are detailed below:
- Exclude opinions where the opinionholder is "Unknown". To remove the unknown opinion holders add opholder=-Unknown to your query, e.g. "sent=Obama&opholder=-Unknown"
- Disambiguate ambiguous search words by adding additional keywords and using boolean search logic. For example, to search for opinions about the movie "The Fighter", it might be helpful to phrase it as sent=Fighter AND (Oscar OR movie).
- Exclude opinions where no topic associated with opinion fails to exist in the opinion sentence, the pre-sentence or post-sentence text. This cannot be done directly as part of the query, but must be applied as post-process filter where the sentence text is checked against the topic list.
- Exclude opinions where the opinion holder is not in the opinion sentence, the pre-sentence or post-sentence text. Like the previous recommendation, this cannot be applied directly through the API but must be applied as a post-process filter.
- Filter out opinions that do not have a doclink or where doclink is empty or resolvable. This is useful for applications that depend on linking to the 3rd party resources referenced by the opinion
- Filter out opinions that do not have an author.