全國中小學科展

四等獎

Desert to Fertile Land: Developing TEPA‐modified montmorillonite clay as an efficient CO2 adsorbent to enhance soil fertility

Global warming is a phenomenon in which the Earth's overall temperature rises as a result of increasing concentrations of greenhouse gases in the atmosphere. Among the major greenhouse gases, carbon dioxide (CO2) is the primary greenhouse gas that contributes significantly to global warming [1,2]. The concentration of carbon dioxide in the atmosphere is rising due to human activities such as burning fossil fuels (coal, oil, natural gas), as well as changes in land use and vegetation [3]. Carbon dioxide and other gases, such as methane and nitrogen monoxide absorb infrared radiation and redirect it back to Earth, warming the planet [4]. This rise in temperature can impact ecosystems, climate, water resources, agriculture, public health, and societies in general [5]. To combat global warming and reduce carbon dioxide concentrations in the atmosphere, many countries around the world, including Saudi Arabia, are working to achieve a vision to reduce carbon emissions by reducing their carbon emissions by 278 million tons per year by 2030 in line with the Paris Agreement, for climate. The Kingdom is committed to generating 50% of its electrical energy from renewable sources by 2030. In addition to the shift in the local energy mix, the Saudi Green Initiative is implementing a number of ambitious programs and projects to reduce emissions. These programs include investing in new energy sources, promoting energy efficiency, and expanding carbon capture and storage programs [6]. Through these initiatives, the Kingdom will be able to achieve its climate goals and establish a sustainable future (Figure 1). In addition, the Paris Climate Change Agreement includes 196 countries and the European Union, covering most of the world. This agreement aims to achieve carbon neutrality by taking measures to reduce carbon dioxide emissions [7].

使用大型語言模型生成音樂中的故事

本研究旨在探索大型語言模型如何應用於音樂生成故事。研究動機源自音樂作為文化中不可或缺的一部分,但若要以文字精準表達出音樂中的故事情緒尚屬困難,藉由本研究提出的方法可以使故事顧及到音樂的情緒起伏。隨著 AI 的發展,我 們開始看到它們在各領域的應用。這項研究的目的是製作出一個系統能以音樂作為輸入,輸出音樂內的故事,為達成目的,我們結合多個模型。研究使用 PyTorch等工具,並探討文句和音樂的共同表示方法,實現情感匹配。研究結果顯示,音樂和文句情感辨識模型表現不錯,也研發出一個完整的生成流程。目前已有直接生成音樂的模型,也有把音樂統整介紹的模型,卻沒有依據音樂中的情緒生成故事的模型。我們研究就是在解決這個問題,結合到 LLaMA2預訓練模型生成出具情緒浮動的故事,要注意的是 LLaMA2的輸出限制最多只能有 4096個token。我們將此產生過程稱為 MTSPL (Music To Story Procedure with LLaMA)。

史坦納樹性質探討與優化

本研究主要目的是想要用數學方法解釋與證明史坦納樹。我們想要知道史坦納樹會滿足怎樣的條件,又或者怎樣的條件不會有最短路徑長。因此從最基本的費馬點出發,到正多邊形、任意多邊形,甚至是正多面體,我們討論各種分布的頂點。在文獻探討與程式模擬中,發現史坦納點的邊數等於三且夾角度數等於一百二十度並非巧合。藉由此性質不斷擴展,可證明出在一般情況下史坦納點的個數會等於頂點的個數減二。有了理論的支持,我們嘗試改進五個柏拉圖多面體的史坦納樹,在正八、十二、二十面體都成功優化成更好的結果。另外,我們也證明出正四面體史坦納樹的結果。有了史坦納樹的幫助,我們可以在不影響城市間連通性的情況下,最大化地縮短道路地總長度,因此可將所需成本最小化。史坦納樹的結果可以代表連通圖的最短路徑長,在工程、建築上都將會是重大的突破。

禽畜糞堆肥除臭與氨氣資源化

禽畜糞堆肥常伴隨氨氣排放而有異味問題。目前成本低、培養易的木黴菌已廣泛用在堆肥中來加速發酵。此外,部分業者會將含氨臭的氣體蒐集再以稀硫酸水洗,產物硫酸銨可作為氮肥,但多被排棄。本研究結合木黴菌添加與排氣稀硫酸水洗,可縮短堆肥期程,更將氨氣肥料化,以 (1)木黴菌合適添加劑量、(2)稀硫酸水洗參數等二項為試驗主軸。結果顯示:(1)添加2/100木黴菌可使堆肥成品中總氮增加13%,減少氨排放;(2)含氨排氣經pH6-7稀硫酸水洗,98%的氨氣轉化成含約1,400mg/L之硫酸銨溶液,氮含量為300mg/L,相當於沼渣沼液農地施灌水準。

In silico Investigation of Cyclosporine Conjugates as Potential Anti-angiogenic Agents via NFAT Inhibition

Calcineurin (CN) activation is a main cause of cancerous tumor formation, one of the leading causes of death globally. Cyclosporine-A (CsA) is a commercially available oral drug that inhibits CN activation; however, low bioavailability limits its use. Nine patented CsA conjugates are potential alternatives to CsA as they have improved cytotoxicities and bioavailabilities but unknown CN-binding affinity. This study aimed to identify the CNinhibition strength and bioavailability of CsA conjugates in silico drug-likeness evaluation via modified Lipinski’s Rule of Five was done on CsA, voclosporin, and CsA conjugates to test bioavailability. The binding affinities of bioavailable compounds were computed via docking to CN in five trials, and the binding affinities were compared. The Water-soluble, RVal, IIA, Alpha, and MeBmt 2 conjugates showed improved bioavailabilities compared to CsA as they passed the drug-likeness screening. After five trials of computational docking to CN, the IIA and RVal conjugates showed improved binding affinities at -15.8 kcal/mol and -15.2 kcal/mol, respectively, compared to CsA at -14.3 kcal/mol. Notably, IIA also showed an improved binding affinity compared to voclosporin at -15.5 kcal/mol. These results suggest that CsA conjugates may be better oral chemotherapeutic drugs than CsA.

修正未切換注音輸入法產生之字元

臺灣的許多人要使用電腦打字時皆會選擇「注音輸入法」作為輸入的方法。然而要使用注音輸入法輸入中文時若無切換輸入法則有可能會誤用到英文輸入法。如要輸入「今天天氣好」五字,使用英文輸入法輸入時會輸出「rup wu0 wu0 fu4cl3」。這種文字不易理解。本研究的目的即為研究將未切換到注音輸入法而打出的英數符號混和字元翻譯為漢字的方法。 本研究使用「PTT 中文語料」與「維基百科中文資料庫」訓練 GRU、BiGRU、LSTM 和 Transformer 以及計算維特比演算法,並與 Google 輸入工具的進行比較。整體來看以 PTT 中文語料計算的維特比演算法的 BLEU4 分數最高,在準確率以及 BLEU4 的評分皆高於Google 輸入工具,分數分別為 0.94 與 88.3 分。 本研究之成果在應用方面極廣,可應用於線上翻譯或聊天軟體的即時翻譯。本研究使用之程式碼開源於 GitHub 頁面,除了可讓使用者下載使用外,使用者也可訓練自己的模型。

布洛卡點的定義擴充與存在之充要條件探討

將平面上A1A2A3的三邊以逆(順)時針方向向內旋轉相同角度後,若三線交於一點,則稱該點為此三角形的正(負)布洛卡點。任意三角形必定同時存在正、負布洛卡點[5],我們的研究主要是推廣布洛卡點的定義到所有的折線多邊形,並進一步研究布洛卡折線多邊形的充要條件與性質。我們定義調和多邊形為正多邊形的反形,並統整出其與特定條件下的圓內接多邊形互為充要的性質,進而發現了圓內接布洛卡多邊形與調和多邊形之間的充要關係,這是目前中外文獻中沒有的。最後,我們試著利用中垂面與垂面將布洛卡點推廣到三維空間。

可同時用於霧水收集與風力發電的石墨烯仿生陣列魔毯

受到地區限制,偏遠地區一直面臨著缺水和缺電的問題。有許多解決方案,其中一個被視為在乾燥地區收集水的有效方法是薄膜霧氣收集技術。這項研究受到自然界沙漠甲蟲的啟發。我們在不吸水的石墨烯/PVDF基底上使用仿生的幾丁聚醣陣列,幫助水滴在霧氣中凝結和滾動脫落,使水收集效率達到0.63LMH。此外,我們使用石墨烯和離子液體一起誘導PVDF晶型自組裝成具壓電性的β相,獲得最適化薄膜的電壓輸出可達到13V(±6.5V)。我們還對於薄膜進行同時取水和取電的可行性評估,結果顯示,在4m/s的霧氣風速下,水收集效率為0.74LMH,發電功率為99.2mW/m2。基於上述研究結果,我們證實了使用單一薄膜利用霧和風作為驅動力,可實現同時產生水和電,這對解決偏遠地區的缺水與缺電問題提供了新的解決方案。

DIVE&CLEAN - Intervention Possible

The DIVE&CLEAN project is an educational and innovative initiative aimed at addressing a significant environmental challenge: marine pollution. With oceans covering over 70% of the Earth’s surface and providing a home to 50–80% of life on the planet, their health is critical. However, marine ecosystems are under threat due to plastic pollution, which impacts wildlife, coastal communities, and global biodiversity. This project centers around the idea of introducing underwater trash bins, especially in areas frequented by recreational divers. While most divers explore the seas without specific tools to collect trash, they could contribute significantly with the right infrastructure. The vision of DIVE&CLEAN is to inspire behavioral change, encourage collaboration, and promote actionable solutions to reduce ocean pollution. Using interactive robotics and storytelling, the project tells the story of divers rescuing animals entangled in plastic and collecting trash from the ocean floor using underwater bins. Through creative performances, it seeks to educate and motivate individuals, resorts, and authorities to adopt sustainable practices.

AI時光機-利用照片轉換技術重溫在地歷史

目前網路上流傳許多使用人工智慧修復照片的網站或應用軟體。然而,由於這些訓練資料多數來自國外,導致修復中式建築照片的效果欠佳。此外,許多老舊照片因氧化、潮濕而泛黃,使得修復程序比起修復純黑白相片更加困難。因此,本研究旨在建🖂一個專門修復中式建築物的機器學習模型,主要分為以下三個部分:首先,使用機器學習模型對老舊照片進行修復,包括著色、去模糊化和降噪;其次,分析使用不同比例之有色調照片(模擬泛黃照片)訓練模型的效果;最後,研究不同的修復順序(著色、去模糊化、降噪)和模型執行次數對照片修復效果的影響,發現「著色、去噪、去模糊化」的順序修復效果最佳。此外,許多老舊照片因為受損等原因,只剩下極少的特徵,因此本研究採用機器學習模型,以延伸重建原始照片。透過這種方法,我們能夠重新建構當時建築物周圍可能的場景和情境。