2016-01-012024-05-18https://scholars.lib.ntu.edu.tw/handle/123456789/697584摘要：我國人口結構逐漸高齡、少子化，家庭原有的養老育幼功能衰退，人們不但需準備退休之後之養老金，在缺少至親或者關係疏離的情況下，甚至得進一步積極交代死亡後的喪儀事項以及遺產處理，例如現實上經由民間公證人公證或認證的遺囑數量，均確實逐年增加。遺囑生效時，遺囑人本人已經死亡，他人無從向本人確認，該貌似遺囑的文書，是否真為遺囑人之意思；故民法對於遺囑有嚴格的形式要求，目的在確保該文書確實是出自於遺囑人之真意。在遺囑人死後，若利害關係人對遺囑效力有爭執，係由法院來判斷其效力。 為了提升遺囑效力的可預測性，讓老年人能安心使用遺囑安排遺產，享受幸福的老年生活，本計畫將分析，何種遺囑（形式）在何種情況（例如：遺囑人年齡、遺囑人健康狀態、遺囑人有無配偶與子女、遺囑之作成有無律師參與等）之下，易被法院認定為有效或無效。由於計畫目標在釐清法院整體裁判趨勢，分析對象不能僅止於單一之裁判，而必須透過大量的分析，始能明瞭。從而，本計畫將嘗試結合人工智慧與法律，進行資訊與法學之合作研究，以達預期目標。在方法上，本計畫將使用資料庫知識發掘系統（Knowledge-Discovery in Database, KDD）中，處理半結構化或無結構資料之文字探勘（Text Mining）技術，探究遺囑當事人之特性、原告之請求、被告之抗辯、主文裁判、裁判基礎條文等項目從統計數據上呈現的特徵，明瞭紛爭的現狀，以及不同時期的裁判是否有不同的特徵。其次，本計畫將對「有效遺囑」與「無效遺囑」定義法律要素，使機器學習自動標記法律要素標籤，進行案件分類（有效遺囑或無效遺囑）以及預測判斷（有效及無效）。 目前為止，法資訊學方面的研究相當稀少。本研究試圖從法學角度投入法資訊學之研究，方法新穎，其成果對法學及人工智慧之研究均有顯著助益。其次，在社會影響力方面，未來有意預立遺囑之高齡者以及欲協助之人（例如：親友、地政士、律師、公證人等），便能趨吉避凶，避免花費精力所為之遺囑卻被認定無效。其次，負責判定遺囑效力之法官，也能藉此明暸其他裁判先例之完整見解與趨向，使所做之裁判更有可預測性與安定性。 <br> Abstract: As the population structure in our country gradually transitions towards one with aging population and low fertility, the traditional family functions including taking care of the elderly and raising young children have retrogressed that people are not only required to prepare for pensions after their retirement but also, taking it one step further, expected to actively prepare and plan for their own funeral etiquette before deaths as well as manage estates. For example, in reality, the number of wills notarized or authenticated by the notaries public is increasing every year as a matter of fact. When a will comes into effect, the testator himself or herself had already died. Other people have no way of confirming with the testator that whether the document resembling a will actually reflects the testator’s intent or not; hence Civil Code has strict requirements for will formality, the goal of which is to ensure that document is actually derived from the testator’s true intent. After the testator dies, if any interested party contests the will’s effects, it is up to the court to determine its effectiveness. To improve the predictability of wills’ effects, making the elderly relieved of utilizing wills to arrange their bequests and enjoy blessed elderly lives, this project will analyze what type of will (formality) would easily be found as valid or invalid with different kinds of factors (for instance: testator’s age, testator’s health condition, whether testator has a spouse or any children). Because the project goal is to clarify the overall trend of court rulings, the subject to analyze cannot be limited to one single judgement. It is necessary to conduct large quantity of analyses in order to understand the trend clearly. Thus, this project will attempt to proceed with combining Artificial Intelligence (A.I.) and Law as an interdisciplinary research for achieving these objectives. Methodologically, this project will utilize a Knowledge-Discovery in Database (K.D.D.) technique, Text Mining to process semi-structured or unstructured information to probe items including testators’ characteristics, plaintiffs’ claims, defendants’ defenses, cases’ holdings, fundamental rules applied, etc. to illustrate their respective features by the statistical data, to understand current statuses of relevant disputes, and whether cases adjudicated in different time periods contain distinct attributes. Secondly, this project will define legal elements for valid and invalid wills, employ machine learning to automatically mark these legal element tags, and conduct categorization of cases (valid or invalid wills) and prediction of decisions (effective and ineffective) based on the tags. There has been fairly few researches on Legal Informatics so far. This research attempts to tackle Legal Informatics from a legal point of view, and employs a novel method. The results would notably benefit researches in both Law and Artificial Intelligence. Furthermore, regarding societal influences, an elderly person who deliberates to make his or her will in advance and any person who is willing to assist him or her with it (for example: relatives or friends, land administration agents, lawyers, notaries public, etc.) are able to count on this approach to Qu ji bi xiong (to seek good luck and avoid calamity) and preclude spending efforts making a will which is adjudicated as invalid. Moreover, judges responsible of probating the validity of wills can also depend on it to understand complete opinions and reasoning of other precedents and their tendencies, thereby improving predictability and stability of the judgements made.法資訊學遺囑法院裁判文字探勘人工智慧legal informaticswillscourt decisionstext miningartificial intelligence