Y, I tracked this down within Solr. This is a feature, not a bug. I
found a solution (set {{captureAttr}} to {{true}}):
https://issues.apache.org/jira/browse/TIKA-2814?focusedCommentId=16745263&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-16745263
Please, though, for the sake of Solr, please run Tika outside of Solr
in production (e.g. SolrJ...see:
https://lucidworks.com/2012/02/14/indexing-with-solrj/)
On Thu, Jan 17, 2019 at 2:15 AM Zheng Lin Edwin Yeo
<edwinye...@gmail.com> wrote:
>
> Based on the discussion in Tika and also on the Jira (TIKA-2814), it was
> said that the issue could be with the Solr's ExtractingRequestHandler, in
> which the HTMLParser is either not being applied, or is somehow not
> stripping the content of <span/> elements. Straight Tika app is able to do
> the right thing.
>
> Regards,
> Edwin
>
> On Tue, 15 Jan 2019 at 10:56, Zheng Lin Edwin Yeo <edwinye...@gmail.com>
> wrote:
>
> > Hi Alex,
> >
> > Thanks for the suggestions.
> > Yes, I have posted it in the Tika mailing list too.
> >
> > Regards,
> > Edwin
> >
> > On Mon, 14 Jan 2019 at 21:16, Alexandre Rafalovitch <arafa...@gmail.com>
> > wrote:
> >
> >> I think asking this question on Tika mailing list may give you better
> >> answers. Then, if the conclusion is that the behavior is configurable,
> >> you can see how to do it in Solr. It may be however, that you need to
> >> do the parsing outside of Solr with standalone Tika. Standalone Tika
> >> is a production advice anyway.
> >>
> >> I would suggest the title be something like "How to prefer plain/text
> >> part of an email message when parsing .eml files".
> >>
> >> Regards,
> >> Alex.
> >>
> >> On Mon, 14 Jan 2019 at 00:20, Zheng Lin Edwin Yeo <edwinye...@gmail.com>
> >> wrote:
> >> >
> >> > Hi,
> >> >
> >> > I have uploaded a sample EML file here:
> >> >
> >> https://drive.google.com/file/d/1z1gujv4SiacFeganLkdb0DhfZsNeGD2a/view?usp=sharing
> >> >
> >> > This is what is indexed in the content:
> >> >
> >> > "content":" font-size: 14pt; font-family: book antiqua,
> >> > palatino, serif; Hi There, <br><br> font-size: 14pt; font-family:
> >> > book antiqua, palatino, serif; My client owns the domain name “
> >> > font-size: 14pt; color: #0000ff; font-family: arial black, sans-serif;
> >> > TravelInsuranceEurope.com font-size: 14pt; font-family: book
> >> > antiqua, palatino, serif; ” and is considering putting it in market.
> >> > It is keyword rich domain with good search volume,adword bidding and
> >> > type-in-traffic. <br><br> font-size: 14pt; font-family: book
> >> > antiqua, palatino, serif; Based on our extensive study, we strongly
> >> > feel that you should consider buying this domain name to improve the
> >> > SEO, Online visibility, brand image, authority and type-in-traffic for
> >> > your business. We also do provide free 1 year hosting and unlimited
> >> > emails along with domain name. <br><br> font-size: 14pt;
> >> > font-family: book antiqua, palatino, serif; Besides this, if you need
> >> > any other domain name, web and app designing services and digital
> >> > marketing services (SEO, PPC and SMO) at reasonable charges, feel free
> >> > to contact us. <br><br> font-size: 14pt; font-family: book antiqua,
> >> > palatino, serif; Best Regards, <br><br> font-size: 14pt;
> >> > font-family: book antiqua, palatino, serif; Josh <br><br>",
> >> >
> >> >
> >> > As you can see, this is taken from the Content-Type: text/html.
> >> > However, the Content-Type: text/plain looks clean, and that is what we
> >> want
> >> > it to be indexed.
> >> >
> >> > How can we configure the Tika in Solr to change the priority to get the
> >> > content from Content-Type: text/plain instead of Content-Type:
> >> text/html?
> >> >
> >> > On Mon, 14 Jan 2019 at 11:18, Zheng Lin Edwin Yeo <edwinye...@gmail.com
> >> >
> >> > wrote:
> >> >
> >> > > Hi,
> >> > >
> >> > > I am using Solr 7.5.0 with Tika 1.18.
> >> > >
> >> > > Currently I am facing a situation during the indexing of EML files,
> >> > > whereby the content is being extracted from the Content-type=text/html
> >> > > instead of Content-type=text/plain.
> >> > >
> >> > > The problem with Content-type=text/html is that it contains alot of
> >> words
> >> > > like "*FONT-SIZE: 9pt; FONT-FAMILY: arial*" in the content, and all of
> >> > > these get indexed in Solr as well, which makes the content very
> >> cluttered,
> >> > > and it also affect the search, as when we search for words like
> >> "font", all
> >> > > the contents gets returned because of this.
> >> > >
> >> > > Would like to enquire on the following:
> >> > > 1. Why Tika didn't get the text part (text/plain). Is there any way to
> >> > > configure the Tika in Solr to change the priority to get the text part
> >> > > (text/plain) instead of html part (text/html).
> >> > > 2. If that is not possible, as you can see, the content is not clean,
> >> > > which is not right. How can we get this to be clean when Tika is
> >> extracting
> >> > > text?
> >> > >
> >> > > Regards,
> >> > > Edwin
> >> > >
> >>
> >
