may contain entity mentions into a working set of documents that should

be more manageable for the subsequent stages.

Nevertheless, this step has a large impact on the recall that can be

maximally attained! Therefore, in this study, we have focused on just

this filtering stage and conduct an in-depth analysis of the main design

decisions here: how to cleans the noisy text obtained online, 

the methods to create entity profiles, the

types of entities of interest, document type, and the grade of

relevance of the document-entity pair under consideration.

We analyze how these factors (and the design choices made in their

corresponding system components) affect filtering performance.

We identify and characterize the relevant documents that do not pass

estimate a practical upper-bound of recall for entity-centric stream

filtering.

\end{abstract}

% A category with the (minimum) three required fields

\category{H.4}{Information Filtering}{Miscellaneous}

%A category including the fourth, optional field follows...

%\category{D.2.8}{Software Engineering}{Metrics}[complexity measures, performance measures]

\terms{Theory}

entity profiling, type of entity filtered for (Wikipedia or Twitter), and

document category (social, news, etc) on the filtering components'

performance. The main contribution of the

paper are an in-depth analysis of the factors that affect entity-based

stream filtering, identifying optimal entity profiles without

compromising precision, describing and classifying relevant documents

that are not amenable to filtering , and estimating the upper-bound

of recall on entity-based filtering.

The rest of the paper  is organized as follows. Section \ref{sec:desc}

describes the dataset and section \ref{sec:fil} defines the task. In

discussion of our method in \ref{sec:mthd}. Following that,  we

present the experimental results in \ref{sec:expr}, and discuss and

analyze them in \ref{sec:analysis}. Towards the end, we discuss the

impact of filtering choices on classification in section

\ref{sec:impact}, examine and categorize unfilterable documents in

section \ref{sec:unfil}. Finally, we present our conclusions in

\ref{sec:conc}.

 \section{Data Description}\label{sec:desc}

We base this analysis on the TREC-KBA 2013 dataset%

\emph{relevant}. Documents with no relevant content are labeled

\emph{neutral} and spam is labeled as \emph{garbage}. 

%The inter-annotator agreement on vital in 2012 was 70\% while in 2013 it

%is 76\%. This is due to the more refined definition of vital and the

%distinction made between vital and relevant.

\subsection{Breakdown of results by document source category}

%The results of the different entity profiles on the raw corpus are

%broken down by source categories and relevance rank% (vital, or

%relevant).  

In total, the dataset contains 24162 unique entity-document

and the remaining 17424 as relevant.

All documents are categorized into 8 source categories: 0.98\%

arxiv(a), 0.034\% classified(c), 0.34\% forum(f), 5.65\% linking(l),

11.53\% mainstream-news(m-n), 18.40\% news(n), 12.93\% social(s) and

50.2\% weblog(w). We have regrouped these source categories into three

groups ``news'', ``social'', and ``other'', for two reasons: 1) some groups

are very similar to each other. Mainstream-news and news are

similar. The reason they exist separately, in the first place,  is

because they were collected from two different sources, by different

groups and at different times. we call them news from now on.  The

same is true with weblog and social, and we call them social from now

on.   2) some groups have so small number of annotations that treating

 through stream of incoming information to find documents that are

 relevant to a set of user needs represented by profiles''

 \cite{robertson2002trec}. Its information needs are long-term and are

 represented by persistent profiles, unlike the traditional search system

 whose adhoc information need is represented by a search

 query. Adaptive Filtering, one task of the filtering track,  starts

 with  a persistent user profile and a very small number of positive

 examples. A filtering system can improve its user profiles with a

 feedback obtained from interaction with users, and thereby improve

 its performance. The  filtering stage of entity-based stream

 filtering and ranking can be likened to the adaptive filtering task

 of the filtering track. The persistent information needs are the KB

 examples.

Stream filtering is then the task to, given a stream of documents of news items, blogs

 and social media on one hand and a set of KB entities on the other,

 to filter the stream for  potentially relevant documents  such that

 the relevance classifier(ranker) achieves as maximum performance as

 possible.  Specifically, we conduct in-depth analysis on the choices

 and factors affecting the cleansing step, the entity-profile

 construction, the document category of the stream items, and the type

 of entities (Wikipedia or Twitter) , and finally their impact overall

 performance of the pipeline. Finally, we conduct manual examination

 of the vital documents that defy filtering. We strive to answer the

 following research questions:

\begin{enumerate}

  \item Does cleansing affect filtering and subsequent performance

  \item What is the most effective way of entity profile representation

  \item Is filtering different for Wikipedia and Twitter entities?

  \item Are some type of documents easily filterable and others not?

  \item Does a gain in recall at filtering step translate to a gain in F-measure at the end of the pipeline?

  \item What characterizes the vital (and relevant) documents that are

    missed in the filtering step?

\end{enumerate}

The TREC filtering and the filtering as part of the entity-centric

TREC filtering track's goal is the binary classification of documents:

relevant or not. In our case, the documents have relevance ranking and

the goal of the filtering stage is to filter as many potentially

relevant documents as possible, but less  irrelevant documents as

as part of the pipeline needs that delicate balance between retrieving

filtering in this case can only be studied by binding it to the later

evaluation.

To achieve this, we use recall percentages in the filtering stage for

the different choices of entity profiles. However, we use the overall

performance to select the best entity profiles.To generate the overall

pipeline performance we use the official TREC KBA evaluation metric

and scripts \cite{frank2013stream} to report max-F, the maximum

F-score obtained over all relevance cut-offs.

\section{Literature Review} \label{sec:lit}

 A more recent manifestation of that is the introduction of TREC KBA in 2012.  Following that, there have been a number of research works done on the topic \cite{frank2012building, ceccarelli2013learning, taneva2013gem, wang2013bit, balog2013multi}.  These works are based on KBA 2012 task and dataset  and they address the whole problem of entity filtering and ranking.  TREC KBA continued in 2013, but the task underwent some changes. The main change between  the 2012 and 2013 are in the number of entities, the type of entities, the corpus and the relevance rankings.

The number of entities increased from 29 to 141, and it included 20 Twitter entities. The TREC KBA 2012 corpus is 1.9TB after xz-compression and has  400M documents. By contrast, the KBA 2013 corpus is 6.45 after XZ-compression and GPG encryption. A version with all-non English documented removed  is 4.5 TB and consists of 1 Billion documents. The 2013 corpus subsumed the 2012 corpus and added others from spinn3r, namely main-stream news, forum, arxiv, classified, reviews and meme-tracker.  A more important difference is, however, a change in the definitions of relevance ratings vital and relevant. While in KBA 2012, a document was judged vital if it has citation-worthy content for a given entity, in 2013 it must have the freshliness, that is the content must trigger an editing of the given entity's KB entry. 

While the tasks of 2012 and 2013 are fundamentally the same, the approaches  varied due  to the size of the corpus. In 2013, all participants used filtering to reduce the size of the big corpus.   They used different ways of filtering: many of them used two or more of different name variants from DBpedia such as labels, names, redirects, birth names, alias, nicknames, same-as and alternative names \cite{wang2013bit,dietzumass,liu2013related, zhangpris}.  Although most of the participants used DBpedia name variants none of them used all the name variants.  A few other participants used bold words in the first paragraph of the Wikipedia entity's profiles and anchor texts from other Wikipedia pages  \cite{bouvierfiltering, niauniversity}. One participant used Boolean \emph{and} built from the tokens of the canonical names \cite{illiotrec2013}.  

All of the studies used filtering as their first step to generate a smaller set of documents. And many systems suffered from poor recall and their system performances were highly affected \cite{frank2012building}. Although  systems  used different entity profiles to filter the stream, and achieved different performance levels, there is no study on and the factors and choices that affect the filtering step itself. Of course filtering has been extensively examined in TREC Filtering \cite{robertson2002trec}. However, those studies were isolated in the sense that they were intended to optimize recall. What we have here is a different scenario. Documents have relevance rating. Thus we want to study filtering in connection to  relevance to the entities and thus can be done by coupling filtering to the later stages of the pipeline. This is new to the best of our knowledge and the TREC KBA problem setting and data-sets offer a good opportunity to examine this aspect of filtering. 

Moreover, there has not been a chance to study at this scale and/or a study into what type of documents defy filtering and why? In this paper, we conduct a manual examination of the documents that are missing and classify them into different categories. We also estimate the general upper bound of recall using the different entities profiles and choose the best profile that results in an increased over all performance as measured by F-measure. 

\section{Method}\label{sec:mthd}

  All Entities    &59.0  &72.2  &79.8  &90.2\\

\hline

\end{tabular}

\end{center}

\label{tab:name}

\end{table}

The upper part of Table \ref{tab:name} shows the recall performances on the cleansed version and the lower part on the raw version. The recall performances for all entity types  are increased substantially in the raw version. Recall increases on Wikipedia entities  vary from 8.2 to 12.8, and in Twitter entities from 6.8 to 26.2. In all entities, it ranges from 8.0 to 13.6.  The recall increases are substantial. To put it into perspective, an 11.8 increase in recall on all entities is a retrieval of 2864 more unique document-entity pairs. %This suggests that cleansing has removed some documents that we could otherwise retrieve. 

\subsection{Entity Profiles}

%\begin{table*}

%\caption{Breakdown of recall percentage increases by document categories }

%\begin{center}\begin{tabular}{l*{9}{c}r}

% && \multicolumn{3}{ c| }{All entities}  & \multicolumn{3}{ c| }{Wikipedia} &\multicolumn{3}{ c| }{Twitter} \\ 

% & &others&news&social & others&news&social &  others&news&social \\

%\hline

% 

%\multirow{4}{*}{Vital}	 &cano-part $-$ cano  	&8.2  &14.9    &12.3           &9.1  &18.6   &14.1             &0      %&0       &0  \\

%                         &all$-$ cano         	&12.6  &19.7    &12.3          &5.5  &15.8   &8.4             &73   &35%.9    &38.3  \\

%	                 &all-part $-$ cano\_part&9.7    &18.7  &12.7       &0    &0.5  &5.1        &93.2 & 93 &64.4 \\%

\paragraph*{LengthBody} Term count of document body

\paragraph*{LengthAnchor} Term count  of document anchor

\paragraph*{FirstPosNorm} 

  Term position of the first occurrence of the target entity in the document 

  body normalised by the document length 

\paragraph*{MentionsBody }

  No. of occurrences of the target entity in the  document body

  In here, we present results showing how  the choices in corpus, entity types, and entity profiles impact these latest stages of the pipeline.  In tables \ref{tab:class-vital} and \ref{tab:class-vital-relevant}, we show the performances in max-F. 

\begin{table*}

\caption{vital performance under different name variants(upper part from cleansed, lower part from raw)}

\begin{center}

\begin{tabular}{ll@{\quad}lllllll}

\hline

%&\multicolumn{1}{l}{\rule{0pt}{12pt}}&\multicolumn{1}{l}{\rule{0pt}{12pt}cano}&\multicolumn{1}{l}{\rule{0pt}{12pt}canonical partial }&\multicolumn{1}{l}{\rule{0pt}{12pt}name-variant }&\multicolumn{1}{l}{\rule{0pt}{50pt}name-variant partial}\\[5pt]

  &&cano&cano-part&all  &all-part \\

   all-entities &max-F& 0.241&0.261&0.259&0.265\\

%	      &SU&0.259  &0.258 &0.263 &0.262 \\	

1) cleansing impacts Twitter

entities and relevant documents.  This  is validated by the

observation that recall  gains in Twitter entities and the relevant

categories in the raw corpus also translate into overall performance

gains. This observation implies that cleansing removes relevant and

social documents than it does vital and news. That it removes relevant

documents more than vital can be explained by the fact that cleansing

removes the related links and adverts which may contain a mention of

the entities. One example we saw was the the cleansing removed an

image with a text of an entity name which was actually relevant. And

that it removes social documents can be explained by the fact that

social. So in both cases social seem to suffer from text

%%%% NEEDS WORK:

Taking both performance (recall at filtering and overall F-score

during evaluation) into account, there is a clear trade-off between

using a richer entity-profile and retrieval of irrelevant

documents. The richer the profile, the more relevant documents it

retrieves, but also the more irrelevant documents. To put it into

perspective, lets compare the number of documents that are retrieved

with  canonical partial and with name-variant partial. Using the raw

corpus, the former retrieves a total of 2547487 documents and achieves

a recall of 72.2\%. By contrast, the later retrieves a total of