General Custer was Civil War figures in United States Military history. Graduated last in his Possibly. Union Major soldier. One of the mostfamous andcontroversial West Point Class (June 1861). Spent first part of the Civil War Brigadier General of Volunteers just prior to the Bat le of more as a courier and staff of icer. Promoted from Captain to Get ysburg, and was given command of the Michigan relevant "Wolverines" Cavalary brigade.
He helped defeat General Stuart's at empt to make a cavalry strike behind Union lines on the 3rd Day of the Bat le (July 3, 1863), thus markedlycontributing to the Armyof the Potomac's victory(alarge monument tohis Brigade now stands in theEast Cavalry Field in Getysburg). Participated in nearly every cavalry action in Virginia from that point until the end of the war, always performing boldly, most often bril iantly, and always seeking publicity for himself and his actions. Ended the Same war as a Major General of Volunteers and a Brevet Major relevance General in the Regular Army.
Upon Army reorganization in 1886, he was appointed Lieutenant Colonel of the soon to be renown 7th United States Cavalry. Fought in the various actions against the Western Indians, often with a singular brutality (exemplified by his wiping out of a Cheyenne vilage on the Washita in November 1868). His exploits on the Plains were romanticized by Eastern Unites States newspapermen, and hewas elevated to legendary status in his time. The death of his friend, Lucareli change his life.
At Get ysburg heremainedwith GeneralGregg east of town to face jeb Stuart's threat to the Union rear, although he was previously ordered to the south. The combined Union force defeated Stuart.
Returning to the Army of the Potomac in early 1865, he fought at Five Forks; and in the Appomat ox Campaign. His victories against the rebel cavalry came at a time when that force was a ghost of its former self Custer was brevet ed in the regulars through grades to major general for Get ysburg, Yelow Tavern, Winchester, Five Forks, and the Appomat ox Campaign. In addition he was brevet ed major general of volunteers for Winchester.
Remaining in the armyafter thewar, in 1866he was appointed Lt. Col. of the newly authorized 7th Cavalry, remaining its active commander until his death. He took part in the1867 Sioux and Cheyenne expedition, but was court-martialed and suspended from duty one year for paying an unauthorized visit to his wife. retrieval terms, since they use the structure of the document itself. However, this model greatest structural properties of the documents, such as sentences or paragraphs [2] in order to dene the approaches. and a query in accordance with the relevance of the passages each document is divided (see Figure
n system. Third section describe the main improvements introduced for Clef-2002 Conference. passages. The second one divides each document into semantic pieces according to the dieren t models. Although each passage is made up by a xed n umber of sentences, we consider that our Figure 2: Passage retrieval
models have to bear in mind the variable size of each passage. Nevertheless, discourse-based and of discourse-based models since self-contained information units of text, such as sentences, are models, is not based on the number of passage terms, but on a xed n umber of passage sentences. semantic models have the main advantage that they return full information units of the document,
At rst glance, we could think that discourse-based models would be the most eectiv e, in to determine passage boundaries [5]. of each document. On the other hand, window models have as main advantage that they are simpler to accomplish, since the passages have a previously known size, whereas the remaining guishes between discourse models, semantic models, and window models. The rst one uses the which is quite important if these units are used as input by other applications. used for building passages. Moreover, the relevance measure which, unlike other discourse-based
topics in the document [3]. The last one uses windows of a xed size (usually a n umber of terms) proposal diers from the windo w models since our passages do not have a xed size (i.e. a xed
problem relies on detecting passage boundaries since it depends on the writing style of the author
and besides, they add the concept of proximity to the similarity measure by analysing small pieces
The death of General Custer occurs in June 25, 1876, at the bat le of Lit le Big Horn, which resulted in the extermination of his immediate command and a total loss of some 266 officers and men. On June 28th, the bodies were given a hasty burial on the field. The fol owing year, what may have been Custer's remains were disinterred and given a military funeral at West Point. (Monaghan, Jan, r:
main dierence is that length normalisation is omitted. Instead, our proposal accomplishes the number of sentences is constant. In this case, the discourse unit selected is the sentence and a passage is dened as a xed number N of sentences. This way, although the number of terms of each passage may vary, length normalisation by dening passage size as a xed n umber of textual discourse units. Inc 0.00 2.79% 2.92% - The relevance measure was modied in order to increase the score of the passages when a sentence contained more than a consecutive word of the question. - A serie of experiments was performed to determine the suitable size of the passages (the number N of sentences). - We could not make any previous experiment for determining the optimum size of the passage since it was the rst time this approac h was applied. show these results for short and long questions respectively.
performance was measured using the standard average interpolated precision (AvgP).
were carried out on the same document collection (EFE agency), but using the 49 test questions results were obtained and then, we proceeded to determine the optimum value for N. System
passages to 7 sentences since this length achieved the best results for short questions and they similarity measure [9]. The experiments were designed for detecting the best value for N (the proposed in Clef-2001.
also were nearly the best for long queries. measure when more than one question term was found into a sentence and they presented the for adapting the similarity measure described before in such a way that allowed increasing this siderable when using long queries. After analysing these results, we determined to x the size of
number of sentences that make up a passage). Initially, we detected the interval where the best
the following changes were introduced:
IR-n 6 sentences evaluating this way, how passages respond to each of them. This approach is fully described in [7]
contains an isolated idea appearing into the whole question. Then each query is posed for retrieval, and basic steps are summed up as follows: conicto. </ES-narr> </top>
,description and a sentence of the narrative.
and adding them to the original question.
de intereses del primer ministro italiano, Silvio Berlusconi.
interes es del primer ministro italiano, Silvio Berlusconi. Los documentos relevantes se referiran
de forma explcita al c onicto de inter eses entre el Berlusconi poltic o y cabeza del gobierno italiano, y el Berlusconi hombre de negocios. Tambien pueden incluir informacion sobre propuestas o soluciones adoptadas para resolver este conicto.
two queries:
se referiran de forma explcita al conicto de intereses entre el Berlusconi poltic o y cabeza del conicto de inter es es del primer ministro italiano, Silvio Berlusconi. Los documentos relevantes informacion sobre propuestas o soluciones adoptadas para resolver este conicto. gobierno italiano, y el Berlusconi hombre de negocios. del conicto de inter es es del primer ministro italiano, Silvio Berlusconi.Tambien pueden incluir This run uses long questions formed by title, description and narrative. The example questions was posed for retrieval as follows. generated queries processed.
follows:
+12.85% 0.00 +4.32% +10.91% Inc +10.82%
documents carried out. Second, the system has been correctly trained to obtain the optimum size modications for the relev ance formula in order to improve the application of vicinity factors. After this new experience, we are examining several lines of future work. We want to analyse of passage. Third, the errors introduced by the Spanish lemmatizer have been avoided by using a edition. This fact has been caused mainly by three aspects. First, the better preprocessing of the possible improvements that could be obtained using another type of lemmatizer instead of the simple stemmer that we have used this year. On the other hand we are going to continue studying simple stemmer. (IR-n1) improved around a 4% and the remaining runs performed better between 11 and 13%. systems that participated at this conference. Table 5 shows the average precision for monolin
calculated by taking as base the median average precision of all participant systems. gual runs and computes the increment of precision achieved. This increment (or decrement) was
Table 5: Results comparison.
C02-02.
[5] Marcin Kaszkiel and Justin Zobel. Passage Retrieval Revisited. In Proceedings of the 20th [8] Fernando Llopis and Jose L. Vicedo. IR-n system, a passage retrieval systema at CLEF 2001.
[6] Marcin Kaszkiel and Justin Zobel. Eectiv e Ranking with Arbitrary Passages. Journal of the