Hi Lance, Thanks for the detailed advice. I was reading Weka menu just now and it did have many classification algorithms. I will start with it and try to follow the two-part process. Will post again if facing difficulties. Thanks again.
On Sun, Jul 10, 2011 at 7:56 AM, Lance Norskog <goks...@gmail.com> wrote: > There is much more to learn about sentiment analysis than about Solr. > I suggest getting one of these toolkits yourself, write some code, and > make some charts. > > Classification is a two-part process: first make a large dataset of > "positive" & "negative" text and train a model to understand the > difference. Second, use the model to evaluate unknown text. The second > part could be added to Solr as an updating process, to include a > positive/negative score with each document. The first part, training > the model, is a batch process done outside. > > The Weka toolkit has more features than Lingpipe, OpenNLP, UIMA etc. I > would start with that. And learn how to program with R. > > Lance > > On Fri, Jul 8, 2011 at 8:13 PM, Zheng Qin <qinzheng...@gmail.com> wrote: > > Thanks, Bruno and Matthew. I saw that tutorial before and Lingpipe > requires > > a license while we are looking at open source solutions. We are not clear > > yet on how to use Solr to do sentiment analysis. Does a NLP or learning > tool > > have to be used to accomplish this task? If a tool is needed, how it can > be > > integrated with Solr? Then, what are the steps? By using classification? > We > > are new to sentiment analysis and any suggestion is welcomed. > > > > On Sat, Jul 9, 2011 at 4:07 AM, Matthew Painter > > <matthew.pain...@kusiri.com>wrote: > > > >> Note you can't use lingpipe commercially without a license though I > >> believe. > >> > >> Sent from my iPhone > >> > >> On 8 Jul 2011, at 18:20, Bruno Adam Osiek <baos...@gmail.com> wrote: > >> > >> > Try Lingpipe. They use Language Models as their engine for sentiment > >> analysis. At (http://alias-i.com/lingpipe/) you will find a > step-by-step > >> tutorial on how to implement it. > >> > > >> > On 07/08/2011 07:14 AM, Zheng Qin wrote: > >> >> Hi, > >> >> > >> >> We are starting a project on Twitter data sentiment analysis. We have > >> >> installed LucidWorks, which also has a Solr admin page. By reading > the > >> >> posts, it seems that sentiment analysis can be done by using OpenNLP > or > >> >> machine learning (Mahout or Weka). Can you share with us which tool > is > >> good > >> >> at classifying positive/negative tweets? Also how to use it together > >> with > >> >> Solr (we only found one posted by Grant on March 16 2010 about > >> integrating > >> >> Solr with Mahout). Your reply will be appreciated. Thanks. > >> >> > >> > > >> > > > > > > -- > Lance Norskog > goks...@gmail.com >
